# Nicholas Zesty Blog > A collection of thoughts, experiments, and adventures in tech, nature, and entrepreneurship by Nicholas Johnson. Public Ghost content for AI and LLM tooling. This file includes a bounded export of public pages first, then recent public posts. Append `.md` to any post or page URL to get the content in Markdown (for example, `/example-post.md`). ## Pages ### About Me URL: https://nicholasjohnson.blog/about/ Last updated: 2026-01-03T20:23:20.000Z About Me Hi, I’m Nicholas Johnson. I’m an electrical engineer and computer scientist by training and a founder by accident, someone who’s spent the last decade learning by building, shipping, breaking things, fixing them, and doing it all again. I didn’t start out trying to collect titles. I started by wanting to solve problems that felt real and consequential. That path took me through energy, mobility, AI, and infrastructure, where I’ve worked as a founding engineer, CTO, and eventually CEO. Along the way, I’ve built whole systems end-to-end—hardware and firmware, cloud platforms, mobile apps, and the messy connective tissue that turns prototypes into products people actually rely on. Most of my credibility comes from doing the work. I’m at my best when things are hard, ambiguous, or moving too slowly—when progress requires rolling up sleeves, making decisions with incomplete information, and pushing through real constraints. I’m comfortable living in the details, taking responsibility, and staying with problems long enough to actually finish them. I’ve also experienced the less glamorous parts of building companies: slow decisions, unclear priorities, team tension, board pressure, and the cost of getting things wrong. Those experiences shaped how I think today. They taught me that momentum isn’t created by clever ideas alone—it comes from sustained execution, ownership, and the willingness to engage directly with uncomfortable problems. This blog is a place for me to write honestly about what I’ve learned—what worked, what didn’t, and how building technology is as much about people and incentives as it is about code and systems. If you’re building something ambitious, or just trying to understand how complex products and teams actually come together in the real world, I hope something here is useful. I’m currently building and experimenting through Zesty Labs, where I work on new ideas at the intersection of software, systems, and energy: [https://zestylabs.org](https://zestylabs.org/?ref=nicholasjohnson.blog) If any of this resonates, I’m always glad to connect. ### Founder Mode URL: https://nicholasjohnson.blog/founder-mode/ Last updated: 2025-08-15T05:22:00.000Z September 2024 At a YC event last week, Brian Chesky gave a talk that everyone who was there will remember. Most founders I talked to afterward said it was the best they'd ever heard. Ron Conway, for the first time in his life, forgot to take notes. I'm not going to try to reproduce it here. Instead, I want to talk about a question it raised. The theme of Brian's talk was that the conventional wisdom about how to run larger companies is mistaken. As Airbnb grew, well-meaning people advised him that he had to run the company in a certain way for it to scale. Their advice could be optimistically summarized as "hire good people and give them room to do their jobs." He followed this advice, and the results were disastrous. So he had to figure out a better way on his own, which he did partly by studying how Steve Jobs ran Apple. So far, it seems to be working. Airbnb's free cash flow margin is now among the best in Silicon Valley. The audience at this event included a lot of the most successful founders we've funded, and one after another said that the same thing had happened to them. They'd been given the same advice about how to run their companies as they grew, but instead of helping their companies, it had damaged them. Why was everyone telling these founders the wrong thing? That was the big mystery to me. And after mulling it over for a bit I figured out the answer: what they were being told was how to run a company you hadn't founded — how to run a company if you're merely a professional manager. But this m.o. is so much less effective that to founders it feels broken. Founders can do things that managers can't, and not doing them feels wrong to them, because it is. In effect, there are two different ways to run a company: founder mode and manager mode. Till now, most people, even in Silicon Valley, have implicitly assumed that scaling a startup meant switching to manager mode. But we can infer the existence of another mode from the dismay of founders who've tried it, and the success of their attempts to escape from it. There are, as far as I know, no books specifically about founder mode. Business schools are unaware of their existence. All we have so far are the experiments of individual founders who've been figuring it out for themselves. But now that we know what we're looking for, we can search for it. I hope that in a few years, founder mode will be as well understood as manager mode. We can already guess at some of the ways it will differ. The way managers are taught to run companies seems to be like modular design in the sense that you treat subtrees of the org chart as black boxes. You tell your direct reports what to do, and it's up to them to figure out how. But you don't get involved in the details of what they do. That would be micromanaging them, which is bad. Hire good people and give them room to do their jobs. Sounds great when it's described that way, doesn't it? Except in practice, judging from the report of founder after founder, what this often turns out to mean is: hire professional fakers and let them drive the company into the ground. One theme I noticed both in Brian's talk and when talking to founders afterward was the idea of being gaslit. Founders feel like they're being gaslit from both sides — by the people telling them they have to run their companies like managers, and by the people working for them when they do. Usually when everyone around you disagrees with you, your default assumption should be that you're mistaken. But this is one of the rare exceptions. VCs who haven't been founders themselves don't know how founders should run companies, and C-level execs, as a class, include some of the most skillful liars in the world. \[[1](https://www.paulgraham.com/foundermode.html?ref=nicholasjohnson.blog#f1n)\] Whatever founder mode consists of, it's clear that it will break the principle that the CEO should engage with the company only through their direct reports. "Skip-level" meetings will become the norm instead of a practice so unusual that there's a name for it. And once you abandon that constraint, there are a vast number of permutations to choose from. For example, Steve Jobs used to run an annual retreat for what he considered the 100 most important people at Apple, and these were not the 100 people highest on the org chart. Can you imagine the force of will it would take to do this at the average company? And yet imagine how valuable such a thing could be. It could make a big company feel like a startup. Steve presumably wouldn't have kept having these retreats if they didn't work. But I've never heard of another company doing this. So is it a good idea, or a bad one? We still don't know. That's how little we know about founder mode. \[[2](https://www.paulgraham.com/foundermode.html?ref=nicholasjohnson.blog#f2n)\] Obviously, founders can't keep running a 2000-person company the way they ran it when it had 20\. There will need to be some degree of delegation. Where the borders of autonomy end up, and how sharp they are, will probably vary from company to company. They'll even vary from time to time within the same company, as managers earn trust. So founder mode will be more complicated than manager mode. But it will also work better. We already know that from the examples of individual founders groping their way toward it. Indeed, another prediction I'll make about founder mode is that once we figure out what it is, we'll find that a number of individual founders were already most of the way there — except that in doing what they did, they were regarded by many as eccentric or worse. \[[3](https://www.paulgraham.com/foundermode.html?ref=nicholasjohnson.blog#f3n)\] Curiously enough, it's an encouraging thought that we still know so little about founder mode. Look at what the founders have already achieved, despite facing a headwind of bad advice. Imagine what they'll do once we can tell them how to run their companies like Steve Jobs instead of John Sculley. ### about-experience URL: https://nicholasjohnson.blog/about-experience/ Last updated: 2026-01-03T20:37:11.000Z Jan 2024 – Present ### Baukunst Collective #### Part Of A Collection Of Creative Technologists Part Of A Collection Of Creative Technologists Advancing The Art Of Building At The Forefront Of Technology And Design. Working With Select Practitioners To Advise Founders, Support Company Building, And Promote Collaborative Learning. [ ](https://baukunst.co/?ref=nicholasjohnson.blog) Jan 2024 – Present ### Orange #### Founder & CEO — Now Board Member Founded and led Orange, a full-stack EV charging company focused on making charging viable at multi-family properties. Built the company from early concept through national deployments, owning product, hardware, software, manufacturing, and go-to-market. Raised capital, built and led the team, navigated regulatory and utility complexity, and scaled the platform into real-world operation. Now serve as a board member. [ ](https://orangecharger.com/?ref=nicholasjohnson.blog) Jan 2017 – Jan 2020 ### LYT #### Founder & CEO — Past Board Member Built an AI-powered cloud-based travel signal priority system using real-time data to minimize bus delays. Created the foundation for traffic management systems that optimize regional traffic signals. Current products are AI-based transit signal priority and emergency preemption without hardware in the field. [ ](https://lyt.ai/?ref=nicholasjohnson.blog) June 2015 – Dec 2016 ### Tesla #### Electrical Engineer Worked on the control and thermal systems team, designing and supporting vehicle programs from development through testing. Owned control and thermal hardware for test setups, collaborated closely with cross-functional teams, and designed thermal systems that were validated and produced at mass production scale. [ ](https://www.tesla.com/?ref=nicholasjohnson.blog) ## Posts ### Agency is not about agents URL: https://nicholasjohnson.blog/agency-is-not-about-agents/ Last updated: 2026-08-02T17:15:06.000Z I am tired of the AI content economy. Every feed right now is full of posts about how to use agents, how to prompt better, how some workflow will change your life, usually illustrated with an AI-generated image and capped with a link to a course. A lot of people are trying to sell you a better way to use AI. Here is the funny part: the people who actually use AI well don't need any of it. They just try things until they figure them out. The models change monthly, so every "definitive guide" is stale by the time it arrives. These posts aren't education. They're clickbait. That's why the real way to learn something is to play, explore, figure it out, and communicate with other people, not pay someone for a course who probably isn't even qualified to tell you what AI can actually do. I say this as someone who has spent a career applying ML/AI technology to real-world applications, not commenting on it. I built an AI-powered transit signal-priority and emergency-vehicle management system at LYT. I patented an AI power-management system for buildings. Right now I'm helping a company build foundation models that derive meaningful inferences from neural signals in plants, which is some of the most interesting applied AI I've seen in years. I run agents for hours at a time in my own work. I am not a skeptic of the technology. I am a skeptic of the theater around it. ## **The tool is not the craftsman** Agency is not about agents. It's about a person's ability to orchestrate software, and that is not new. It's what we've been doing for decades. I like to joke that the best engineers are lazy; they will go to great lengths to make things easier. Mastery of Premiere versus fumbling through Final Cut can be the difference between a six-day edit and a three-week one. Give someone who has never touched wood a fully equipped woodshop, and they'll produce junk for a long time. Give a woodworker with ten years of experience that same shop and they'll build something remarkable in a day. The shop was never the point. The tool isn't amazing. The craftsman is. AI is the best-equipped woodshop ever built, handed to everyone at once. That's exactly why the output of most people using it looks sloppy. ## **What the tool actually does** In my own work, roughly half of what AI hands me is wrong in some way. If I weren't a domain expert, or didn't spend the time cross-referencing and actually reading the sources it cites, I would be publishing the same confident nonsense I'm complaining about. Take fundraising, since "let my AI team fundraise for you" is a pitch I keep seeing, usually from people who have never raised a dollar. Raising money is not about automated emails landing in inboxes. It's warm introductions and relationships. If you can't get a warm intro to a top investor, you are probably not getting a check, even with exceptional metrics. Anyone who believes AI has fundamentally changed that doesn't understand how trust works between humans. A CEO recently argued on a podcast that the future belongs to 10 times as many people who can multiply their output a hundredfold with AI. I think that's roughly right and describes the idea of a 10x engineer. But notice what it requires: the foundational knowledge and experience have to already be there. AI doesn't make me smarter. It clears the busywork so my existing judgment goes further. It multiplies output for people qualified to filter it, and it multiplies garbage for everyone else. Same tool, opposite results. ## **We have seen this movie** I've watched more than ten years of self-driving car promises, built by genuinely brilliant engineers grinding on some of the hardest software and hardware problems on the planet. The research goes back decades, to labs in Europe and at UC Berkeley, and we are still only on the precipice. Transformative technology is real, and it is also slow. Those two facts coexist. The same will likely be true for AGI: I have yet to see anything that suggests sentience. What we have is a statistical model, an enormous file of weights mapping inputs to outputs through gradient descent. Extraordinarily useful. Not a mind. My hunch is that anything genuinely close to general intelligence will look more biological than anything we're building today. That is a longer conversation for another post. A friend of mine put it well recently: innovation theater has its pros and cons. As a founder, I understand the value of making noise to get attention. It attracts talent, capital, and customers, all needed for companies to grow. But there's a difference between making noise about what you're building and lying about what the tool can do to sell a lie or a half-hearted course on A.I. to capture people who are afraid they are being left behind. The wild part is that AI has a bad wrap because most people worldwide are afraid it will replace them. ## **The layoff excuse** Which brings me to the claim that AI is replacing jobs. Mostly, it isn't. It's a convenient explanation for layoffs that were already going to happen, one that lets executives skip the harder conversation about why. If AI were genuinely the driver, we'd be seeing it in productivity data before we saw it in press releases. What AI actually does for most workers is simpler and less flattering: it increases the volume of what they produce without improving the quality. More documents, more code, more posts, more slop. That's the pattern underneath every hype cycle, but it takes a long time, lots of hard work, and attention to detail for everything to come together to build truly lasting impact at scale. So let's talk about AI the way we talk about every transformative tool in history, from software to the internet: honestly. It is a genuine step change, similar to a compiler but way more generalized. I won't discredit that. But it rewards people who already know what they're doing in specific niches, and it exposes everyone else trying to find a shortcut. For what it's worth, I used AI to help edit this essay. That's the whole point. The tool helped me say this faster. It didn't decide what was worth saying. Volume is not quality, and quality is not output; there may be a point in time when physical agents and AI will make us reconsider society as a whole, and that is teh more interesting conversation. ### Rivian's 🔥 Burning Cash to Scale URL: https://nicholasjohnson.blog/rivains-burning-up-tesla/ Last updated: 2026-08-25T03:42:27.000Z A white paper on the only American automaker running Tesla's playbook against Tesla, with every material figure sourced and rated. All financial data is drawn from SEC filings, company disclosures, and rated third-party sources; see Methodology and the Bibliography at the end. ## **Executive Summary** Rivian just did the hardest thing in the car business twice. It survived launching a vehicle program from scratch, and now it is launching a second, cheaper one while still losing money on the first. As of August 2026, the scoreboard reads like this: R2 customer deliveries began June 9, 2026, with a trim ladder running from $57,990 today down to $44,990 by summer 2027 \[8\], Q2 2026 deliveries of 12,194 beat the company's own guidance \[4\], full-year guidance was raised to 65,000 to 70,000 vehicles \[4\], and consolidated gross margin hit 11 percent, an all-time high \[6\]. Rivian has consumed roughly $28.2 billion of cumulative losses to get here \[6\], against Tesla's $4.97 billion to reach the equivalent Model 3 milestone in 2017 \[20\], a gap of about 4.4x after adjusting for inflation. This paper argues five things: 1. **Rivian is the only Western automaker taking Tesla head-on with Tesla's own strategy**: vertical integration of software, electronics, compute, and now custom AI silicon, funded by a consumer vehicle business. Everyone else either buys the stack from suppliers or has retreated from EVs entirely. 2. **The capital gap between Rivian and 2017 Tesla is real but explainable**, and the more interesting question is what Rivian bought with the money: a second platform with half the bill of materials, a software business Volkswagen is paying up to $5.8 billion to access, and an autonomy stack credible enough that Uber committed to buy 10,000 robotaxis. 3. **Rivian's autonomy program is a Tesla playbook clone and that is a good thing,** In-house chips, end-to-end learned driving, and a customer-fleet data flywheel, but with radar and, from late 2026, they are looking at adding llidar as ground truth rather than cameras alone. It’s to be seen can the scale software need to compete. 4. **If the autonomy bet lands, the consumer fleet becomes an option on a robotaxi fleet**, which is exactly the thesis Tesla has been selling shareholders for a decade, except Rivian has a signed launch customer. That changes the life time value of there vehicles masisvly. 5. **The demand environment is brutal and the brand environment is Rivian's gift.** The federal tax credit is gone, US EV share has fallen back to 5.8 percent, and Tesla's brand has been damaged badly enough that a majority of Americans say they would not consider one. Someone will absorb the buyers Tesla is shedding. The counterweights that need to be address: Rivian still loses money on every vehicle at the automotive gross profit line as of Q2 2026, guidance implies roughly $2 billion of negative adjusted EBITDA this year from their own projections, the eyes-off autonomy features are announced rather than shipped, and the Georgia plant that carries most of the volume story does not produce until late 2028\. Rivina is a leveraged bet on execution based on where they are today, not a settled outcome. ## **1\. The Price of Admission: What It Cost to Reach a Mass-Market EV** "Money required to put a mass-market EV into production" does not map onto any single accounting line, so this analysis uses the most defensible proxy available: cumulative net loss since inception (the accumulated deficit) at the close of each company's mass-market launch year, with each year's loss inflation-adjusted to 2026 dollars via CPI-U before summing. Full caveats are in the Methodology section. The anchor numbers, straight from the 10-Ks: Capital consumed to reach a mass-market EV ## Tesla Model 3 vs Rivian R2 Cumulative net losses from founding to the launch of each company's first mass-market vehicle — the total money the business burned to get there. Toggle to inflation-adjust. As reported 2026 dollars CPI-U Tesla to Model 3 · 2003–2017 $0.00B Rivian to R2 · 2009–2026 $0.00B 0.0× the capital Rivian consumed to put a mass-market EV into production, relative to Tesla — **in constant 2026 dollars**. Cumulative burn by company age Accumulated deficit plotted against years since each company was founded. Dots mark the mass-market launch. Tesla Rivian TESLA · 14 yrs to Model 3 $0.00B Accumulated deficit, FY2017 year-end RIVIAN · \~16 yrs to R2 $0.00B Accumulated deficit, FY2025 year-end **Method.** "Capital consumed" = accumulated deficit (cumulative net loss since inception) at the close of each company's launch year — FY2017 for Tesla, FY2025 for Rivian. In 2026-dollar mode, each year's incremental loss is reflated to 2026 dollars with CPI-U (2026 base ≈ 327, Feb 2026) before summing, so older Tesla losses are weighted up more than recent Rivian losses. This is total money the business burned, **not** the engineering cost of the car alone — both figures include other products, energy/software lines, and (for Tesla) SolarCity. Tesla's figure includes a half-year of post-launch Model 3 ramp, while Rivian's is measured just before first R2 customer deliveries. Sources: Tesla & Rivian SEC 10-K filings; U.S. BLS CPI-U. Inflation explains part of the apparent gap. Tesla burned 2003 to 2017 dollars; Rivian burned 2020s dollars. But adjusting for that only compresses the ratio from 5.4x to roughly 4.4x. Rivian's capital intensity to reach a mass-market EV is real, not a dollar-vintage illusion. For a third data point: Lucid, which has not yet launched a mass-market vehicle, reported an accumulated deficit of $13.3 billion at the end of 2024 \[82\] and roughly $15.6 billion at the end of 2025 per its FY2025 reporting \[83\]. The lesson generalizes. Nobody gets into this business for less than ten billion dollars anymore. --- ### **Why did it take 4.4x the money?** Four structural reasons, in rough order of importance: **First, Rivian launched into production hell with three vehicles at once.** Tesla ramped one vehicle at a time basily. Rivian launched the R1T pickup, R1S SUV, and the Amazon commercial van within a single three-month window in late 2021, into the worst supply chain environment in modern automotive history. RJ Scaringe made this point directly on the Q2 2026 earnings call: "In R1, we didn't just launch R1, we launched R1T, R1S, and a commercial van all within the same three-month window. In sharp contrast to that the R2, we have a very limited set of build combinations. That was highly intentional to facilitate a smoother and faster ramp" \[12\]. **Second, factories are staggeringly expensive and Rivian paid for capacity ahead of revenue.** Rivian bought the former Mitsubishi plant in Normal, Illinois for just $16 million in January 2017 \[86\], but had invested roughly $750 million into it by early 2020, then committed another $1.5 billion to the R2 expansion that took the site to 215,000 units of annual capacity \[1\]\[17\]. The Georgia plant is a $5 billion commitment \[18\]. For calibration, Tesla's paid in Mya 2010 $42 million in cash to buy the shuttered NUMMI plant (now the Fremont Factory) from Toyota and General Motors' bankrupt spinoff. Concurrently, Toyota agreed to invest $50 million in Tesla's upcoming stock, and Tesla was backed by a $465 million conditional loan from the Department of Energy. During Model 3 ramp Tesla spent $2B on factory upgrades to get the MUMMI factory to 500k vehicles a year in production. Also, note that Tesla Shanghai Gigafactory, the fastest large greenfield auto plant ever built, cost roughly $2 billion for 250,000 units of initial capacity and went from groundbreaking to production cars in under twelve months \[85\]. That is the benchmark Rivian is being measured against, and nobody else has matched it the history of automotive. **Third, vertical integration is expensive before it is cheap.** Rivian developed its own zonal electrical architecture, its own ECUs, its own software stack, its own camera hardware, and now its own AI silicon. The Gen 2 R1 refresh cut the vehicle's ECU count from 17 to 7 and removed 1.6 miles of wiring per vehicle \[37\]. That engineering cost billions before it saved a dollar, but it is exactly the asset Volkswagen is now paying up to $5.8 billion to access \[43\], which retroactively reprices all of that spending from "burn" to "product development." **Fourth, the era subsidized Tesla and taxed Rivian.** Tesla scaled through a decade of near-zero interest rates, an uncontested EV market, rising regulatory credit revenue, and a $465 million DOE loan it repaid nine years early \[27\]. It sold every car it could make. Rivian scaled through pandemic-era supply chains, 2022 to 2023 rate shocks, tariff costs (Q2 2026 automotive results include an IEEPA tariff refund receivable, meaning tariffs were material enough to matter \[6\]), and then the abrupt removal of the $7,500 federal credit in September 2025 \[55\]. One important caveat cuts both ways: accumulated deficit measures the total money each company consumed, not the standalone cost of one car program. Tesla's $4.97 billion includes Roadster, Model S, Model X, the energy business, Supercharge network, and the SolarCity acquisition's losses. Rivian's $27 billion includes R1, the Amazon van program, adventure charge network, the software organization that VW is now paying for, and the autonomy program. Neither figure is "the cost of the Model 3" or "the cost of the R2." They are the cost of the company that could build them. --- ## **2\. R2 by the Numbers: Where Rivian Stands in August 2026** Everything in this section is from Rivian's SEC-filed disclosures unless otherwise noted. ### **Production and deliveries** | Period | Produced | Delivered | Notes | | ------- | -------- | --------- | ------------------------------------------------------------------------------------------------------ | | FY 2025 | 42,284 | 42,247 | Pre-R2 baseline \[2\] | | Q1 2026 | 10,236 | 10,365 | Down \~30% YoY during the R2 line changeover; a tornado also disrupted the Normal facility \[3\]\[11\] | | Q2 2026 | 12,613 | 12,194 | Beat guidance of 9,000 to 11,000; first R2 customer deliveries June 9 \[4\]\[8\] | Full-year 2026 guidance started at 62,000 to 67,000 deliveries in February \[2\], was reaffirmed twice, and was raised to 65,000 to 70,000 on July 2 after the Q2 beat \[4\]. Management expects deliveries to be heavily weighted to Q4 as R2 ramps \[6\]. ### **R2 pricing and positioning** R2 was unveiled in March 2024 with a promised starting price of "around $45,000" \[7\]. The actual ladder, announced with first deliveries on June 9, 2026 \[8\]: - **R2 Performance with Launch Package: $57,990**, available now. Dual-motor AWD, 656 hp, 0 to 60 in 3.6 seconds, EPA-estimated range up to 330 miles, and the Autonomy+ software tier included. - **R2 Premium: $53,990**, late 2026. - **R2 Standard (RWD Long Range): $48,490**, early 2027. - **R2 Standard: $44,990**, summer 2027. So the $45,000 promise survives, but it arrives a year after launch, which is the standard playbook (Tesla ran Model 3 the same way, shipping $55,000+ configurations first and the $35,000 car much later). Early demand signals are strong: a record 57,000+ R2 demo drives in Q2, and Scaringe said reservation-to-order conversion "has been meaningfully higher than what we expected," even at the $58,000 launch price \[6\]\[12\]. ### **The cost structure argument** This is the single most important forward-looking claim in the Rivian story, so it deserves the exact quote. Scaringe, Q1 2026 earnings call: "For R2, our bill of materials is expected to be approximately half of our R1 platform. For non-BOM cost of goods sold, we expect to see a reduction of more than 50%, resulting from a focus on design for manufacturing and leveraging fixed cost efficiencies through higher production volumes" \[11\]. ### **Financial position (Q2 2026, reported July 30, 2026) \[6\]** - Revenue: **$1.658 billion**, up 27 percent year over year. Software and services contributed $515 million of that, up 37 percent, at a 42 percent gross margin. - Consolidated gross profit: **$179 million (11 percent margin)**, a $385 million improvement year over year. Automotive gross profit was still negative at **negative $36 million**, absorbing roughly $100 million of incremental R2 ramp costs. - Net loss: **$837 million**. Adjusted EBITDA: **negative $379 million**, improved from negative $667 million a year earlier. - Cash, equivalents, and short-term investments: **$5.31 billion**, with pro forma liquidity of roughly **$7.2 billion** including the July equity offering (\~$1.3 billion net) and remaining credit facilities \[6\]\[10\]. - Full-year guidance: adjusted EBITDA of negative $2.0 to negative $1.8 billion; capex trimmed to $1.70 to $1.80 billion \[6\]. Management's stated targets: R2 production moves from one shift to two by the end of Q3 2026, and both R2 and total automotive gross profit turn positive "as an exit rate" for 2026 \[12\]\[6\]. The company's stated "North Star" is profitably delivering 4,000 vehicles per week out of Normal \[12\]. For context, 4,000 a week is roughly 200,000 a year, approaching Tesla's total 2018 output during the Model 3 ramp \[91\]. ### **Capacity: the 515,000-unit skeleton** - **Normal, Illinois: 215,000 units per year** across R1T, R1S, the commercial van, and R2 \[1\]\[7\]. - **Stanton Springs, Georgia:** vertical construction began spring 2026, first production late 2028\. The DOE loan was restructured in April 2026 from $6.57 billion down to $4.5 billion, while the initial phase capacity was raised 50 percent to 300,000 units per year \[9\]\[16\]. Georgia builds R2, the R2 robotaxi variant, and R3 \[6\]. - Combined planned footprint: roughly **515,000 units per year**, per management on the Q1 2026 call \[11\]. A note I honesty this moves the economics in both directions: **restructuring a government loan downward** while claiming a **capacity increase** is the kind of thing that deserves scrutiny. The mechanics (a smaller initial building scope, higher planned line rates) are plausible and the DOE signed off, but Georgia is a 2028 story and 2028 stories from EV startups have a poor base rate. It is appropriate to treat 515,000 units as a design target, not a forecast. --- ## **3\. The Expected Growth Curve** ## The Ramp: Tesla vs. Rivian, Aligned by Company Phase Annual vehicle deliveries (a close proxy for production; the two track within a few percent for both companies). Year 0 is each company's mass-market launch year: Model 3 in 2017, R2 in 2026\. By coincidence of history, both companies delivered their first flagship exactly five years earlier (Model S, 2012; R1T, 2021). Hover any point for the calendar year. Rivian's 2026 point is company guidance, not an actual. Toyota's global total (all brands) is shown for scale; the chart defaults to a log scale so all three are readable at once. Aligned by phase Calendar year Linear Log scale Dashed gold segment: Rivian FY2026 guidance midpoint of 67,500 (range 65,000–70,000, raised July 2, 2026). Dotted reference lines: Normal, IL plant capacity (215,000/yr) and Normal plus the Georgia plant's initial phase (515,000/yr combined; Georgia begins production late 2028). Capacity is a ceiling, not a forecast. Toyota appears as a full annual series in calendar view and as a flat reference line (its record 2025 total, 11.32M) in the phase-aligned view, since a 90-year-old incumbent has no comparable launch phase to align. Data table and sources | Phase year | Tesla (cal. year) | Tesla deliveries | Rivian (cal. year) | Rivian deliveries | Toyota global (Tesla cal. year) | | ---------- | ----------------- | ---------------- | ------------------ | ----------------- | ------------------------------- | **Sources** (validity rating in parentheses, per the white paper's 1–5 scale): Tesla annual production & delivery press releases and 8-K exhibits, 2012–2025, via [Tesla IR](https://ir.tesla.com/press-releases?ref=nicholasjohnson.blog) and [SEC EDGAR](https://www.sec.gov/Archives/edgar/data/0001318605/000119312516418701/d106308dex991.htm?ref=nicholasjohnson.blog) (5). Rivian production & delivery 8-K exhibits, 2021–2025, via [SEC EDGAR](https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001874178&type=8-K&ref=nicholasjohnson.blog) (5). Rivian FY2026 guidance: [Q2 2026 delivery 8-K, Jul 2, 2026](https://www.sec.gov/Archives/edgar/data/1874178/000187417826000048/ex-9912q26deliveryproducti.htm?ref=nicholasjohnson.blog) (5). Normal capacity: Rivian FY2025 10-K (5). Georgia initial capacity: [Rivian press release, Apr 30, 2026](https://www.businesswire.com/news/home/20260430750054/en/?ref=nicholasjohnson.blog) (4). Toyota global sales (incl. Lexus, Daihatsu, Hino): [Toyota Motor Corp. official production/sales results](https://global.toyota/en/company/profile/production-sales-figures/202512.html?ref=nicholasjohnson.blog) (4); 2023 (11,233,039) and 2025 (11,322,575, record) are exact as reported; other years rounded to the nearest 10,000; 2024 shown as 10.82M per Toyota's reported 3.7% decline (some trackers list \~11.0M on a different counting basis). Note: Tesla's 2012 figure (\~2,650) is approximate (Model S launched June 2012); Tesla's initially reported 2015 figure (50,580) was later restated as 50,658 in an SEC comment-letter response; the as-reported figure is used. Plotting the disclosed guidance and capacity milestones produces a curve with three distinct phases: **Phase 1 (2026): the changeover year.** 65,000 to 70,000 deliveries \[4\], which is meaningful growth over 2025's 42,247 \[2\] but still a rounding error against Tesla's 1.64 million 2025 deliveries \[22\]. The year's real deliverable is not volume; it is proving that R2 ramps on schedule, converts reservations, and exits the year with positive automotive gross profit \[6\]. **Phase 2 (2027 to mid-2028): the Normal saturation phase.** With R2 Standard trims arriving through 2027 at $44,990 to $48,490 \[8\] and the plant's 215,000-unit ceiling \[1\], the achievable range is bounded. If Rivian approaches its 4,000-per-week North Star \[12\], annualized output approaches 200,000\. Demand at those prices is the open question, examined in Section 8. **Phase 3 (late 2028 onward): the Georgia unlock.** 300,000 initial units of capacity \[9\], carrying R2 volume, the Uber robotaxi build (up to 50,000 units beginning late 2028 \[9\]), and R3, which Rivian has committed to price below R2 but has not dated or priced officially \[7\]. Any specific R3 price you have read is a guess; the company has not published one. For calibration against the incumbent: Tesla delivered roughly 103,000 vehicles in 2017, 245,000 in 2018, 368,000 in 2019, and 500,000 in 2020 during its equivalent phase \[91\]. Rivian's guided 2026 (65,000 to 70,000) resembles Tesla's 2016\. On pure volume, my estimate of a four-to-five-year gap holds up against the disclosed numbers, and it is an estimate, not a company figure. The gap closes only if R2 demand holds at post-subsidy prices and Georgia executes. It widens if either fails. --- ## **4\. The Technology Bet: In-House Silicon and the Real Moat** The reason to write about Rivian in 2026 is not the delivery numbers. It is that on December 11, 2025, at its first Autonomy & AI Day in Palo Alto, Rivian became the second Western automaker in history to take its autonomy stack vertically integrated all the way down to custom silicon, with the hardware slated to ship on R2 beginning late 2026 \[14\]. ### **What Rivian announced, precisely** - **RAP1 (Rivian Autonomy Processor):** a proprietary, purpose-built AI inference chip, developed in collaboration with Arm on the Armv9 architecture using Cortex-A720AE automotive cores \[33\]. Secondary reporting describes it as a 5nm TSMC-fabbed part \[34\]; Rivian itself has not confirmed the foundry, so treat the node as reported rather than official. - **ACM3, the third-generation autonomy computer:** 1,600 sparse INT8 TOPS, processing 5 billion pixels per second \[14\]. - **A "Large Driving Model" (LDM):** Rivian's end-to-end learned driving model with what the company describes as an LLM-like architecture, trained on a data flywheel from the customer fleet. Flagged driving instances upload automatically to Rivian's cloud, are auto-labeled, and reinforcement learning distills improved policies back to the onboard models \[14\]. - **Lidar, added to R2 beginning late 2026,** with a stated purpose worth quoting because it is strategically clever: making "our R2 fleet a very large ground truth fleet for training our model" \[14\]. - **The stated goal, in Scaringe's words:** "Our updated hardware platform, which includes our in-house 1600 sparse TOPS inference chip, will enable us to achieve dramatic progress in self-driving to ultimately deliver on our goal of delivering L4" \[35\]. The shipped product today is more modest, and the distinction matters. Gen 2 R1 vehicles carry 11 exterior cameras, five radars, and a compute module delivering roughly 200 to 250 TOPS (Rivian's own pages disagree on the exact figure, so I am citing the range) \[15\]; trade press identifies the processors as dual NVIDIA Drive Orin, a detail Rivian's own materials do not confirm \[92\]. Hands-free highway driving shipped over the air in March 2025 \[13\]. The paid tier, Rivian Autonomy+, launched at $2,500 one-time or $49.99 per month, with Universal Hands-Free covering 3.5 million miles of North American roads \[15\]. Two honest limitations: Universal Hands-Free does not stop for traffic lights or stop signs \[15\], and the eyes-off features remain roadmap items as of this writing, not shipped software. Rivian said in March 2025 that a hands-off, eyes-off feature was planned for 2026 in controlled conditions \[13\]; as of August 2026, I can find no evidence it has shipped. ### **Why this is Tesla's playbook, amended rather than copied** Here I want to be more precise than my own first draft of this argument. It is tempting to say Rivian is "copying Tesla's stack," and directionally that is right: an in-house inference chip (Tesla did HW1 through AI5), an end-to-end learned driving model (Tesla's FSD v12 onward), and a customer fleet as a data flywheel (Tesla's core structural advantage since 2016). Scaringe even describes it in Tesla vocabulary: an architecture designed "around an AI-centric approach where the data flywheel of our deployed fleet helps make the model better and better through reinforcement learning" \[36\]. But Rivian diverges from Tesla on the most contested question in the field. Tesla is dogmatically vision-only. Rivian runs early sensor fusion across cameras and radar today, and adds lidar on R2 from late 2026, not as a crutch for the driving policy but as a ground-truth training signal for the camera-first model \[13\]\[14\]. That is a hedge Tesla has refused on principle, and it means Rivian gets fleet-scale lidar ground truth that, among Western companies, only Waymo otherwise has, except Waymo's fleet is around 3,000 vehicles \[90\] and Rivian's R2 fleet should pass that within months of the lidar hardware shipping. ### **Why almost nobody else can do this** The claim that "Rivian is one of the only carmakers capable of this" sounds like fan fiction until you enumerate the alternatives, so let us enumerate them. - **GM Super Cruise** is an excellent geofenced L2 system, and it is structurally dependent on pre-built HD maps (roughly 750,000 lidar-mapped miles) and supplier silicon. It does not learn from its fleet end to end. - **Ford BlueCruise** is built around Mobileye's EyeQ and mapping stack \[41\]. Ford does not own the perception model, the chip, or the training loop. When Consumer Reports ranked hands-free systems in October 2023, BlueCruise came first on usability \[39\], which is genuinely to Ford's credit, and it changes nothing structurally: Ford cannot iterate what it does not own. - **The German incumbents** buy from NVIDIA, Qualcomm, and Mobileye in various combinations, and Volkswagen's decision to pay Rivian up to $5.8 billion for electrical architecture and software \[43\] is about as loud an admission as the industry will ever produce that the in-house path failed internally. - **Waymo** owns its full stack and is years ahead on driverless operation, but it is not an automaker; its sixth-generation sensor suite (13 cameras, 4 lidars, 6 radars \[38\]) rides on other companies' vehicles at a reported hardware cost near $20,000 per vehicle (Waymo has never published this number; all such figures are press estimates), and its roughly 3,000-vehicle fleet \[90\] cannot generate consumer-fleet-scale training data. - **The honest Chinese exception:** BYD is running the same vertically integrated play at ludicrous scale, shipping its God's Eye ADAS at no extra cost across its lineup, including lidar-equipped trims on cars as cheap as $13,000 \[42\]\[62\]. The correct framing is that Rivian is the only company doing this *among Western automakers*. China has several, and Section 7 deals with what that means. Building the vehicle from the ground up, with the electrical architecture, the compute, the sensors, and the software designed as one system, is the qualifying condition for this kind of autonomy program. Rivian designed for it from the start, which is why a company that delivered 42,000 vehicles last year can credibly attempt what companies delivering millions cannot. That is the moat. It is not that Rivian's model is better than Tesla's today (there is no evidence it is), but that Rivian is one of exactly two Western car companies structurally *permitted* to play this game, and the other one is busy setting its brand on fire. ### **Where Tesla actually stands, for fairness** Tesla's program remains formidable and ahead on deployment. FSD v14 shipped in October 2025 \[30\], the AI5 chip taped out in April 2026 with Musk claiming up to 40x improvement over AI4 on some metrics \[29\], and the robotaxi service that launched in Austin in June 2025 has expanded to multiple cities \[32\]. But the program has visible cracks: Tesla shut down its Dojo training supercomputer in August 2025 and disbanded the team \[28\], confirmed on the Q1 2026 call that HW3 vehicles cannot achieve unsupervised FSD \[30\], and independent tracking of the robotaxi fleet's early months found a small fleet with heavy operational scaffolding; Electrek's analysis of NHTSA filings implied a crash rate around one per 55,000 miles in the early Austin deployment \[31\]. I rate that source a 3 and note its critical stance toward Tesla, but the underlying filings are federal. Musk says robotaxis will be "widespread in the US by end of 2026" \[32\]. He has made adjacent predictions annually since 2016. --- ## **5\. From Consumer Fleet to Robotaxi Fleet: The Option Value** The R2 thesis has a second act that did not exist a year ago, and it came with a signed contract. In March 2026, Rivian and Uber announced a partnership under which Uber and its fleet partners will purchase **10,000 fully autonomous R2 robotaxis, with an option for up to 40,000 more in 2030**, and Uber will invest **up to $1.25 billion in Rivian through 2031** \[5\]. The robotaxi variant of R2 will be built at the Georgia plant, up to 50,000 units, beginning late 2028 \[9\]. This is the structural difference between Rivian's autonomy story and Tesla's: Tesla's robotaxi economics accrue to Tesla's own network, which requires Tesla to build, own or coordinate, insure, clean, charge, and dispatch a fleet. Rivian sold the fleet problem to the world's largest ride-hail network and kept the hardware and software margin. Reasonable people can argue which captures more value; only one of them is a signed purchase agreement. If personal L4 arrives on R2-class hardware, the consumer math gets strange in a way worth spelling out. A vehicle that can earn ride-hail revenue when idle is not a depreciating asset in the usual sense, and a consumer fleet of a few hundred thousand R2s becomes convertible, at the owner's option, into supply for autonomous networks. Tesla has pitched exactly this to shareholders since 2019\. Rivian has never promised it publicly, which given the industry's track record on autonomy promises may itself be a point in its favor. --- ## **6\. What Tesla Got Right: A Model 3 Retrospective and the Ramp Philosophy** Back in 2015, I wrote a college essay arguing that the Model 3 would be the most revolutionary car of my generation. I cannot find the essay anymore, which is probably for the best, because I was right for the wrong reasons. I thought the revolution would be electrification itself: the $35,000 EV for everyone. What actually happened is that the Model 3's platform spawned the Model Y, and the Model Y became, per JATO Dynamics, the best-selling car in the world in 2023 at 1.22 million units, the first EV ever to hold that title \[63\]. The buyers were not, in the main, buying an electric car. They were buying the software, the driving experience, the over-the-air updates, the minimalist interior that made everything else feel like a rental counter, and the car happened to be electric. Every survey and every conversation I have had with owners points the same direction: "fully electric" is a feature people approve of; the technology experience is what they love. Other automakers have spent a decade trying to build "their Tesla" and mostly shipped compliance vehicles with laggy tablets. Precision requires an update to the sales claim, because the crown has since gotten complicated. Per JATO's data, the Toyota RAV4 retook the global title in 2024 by fewer than 3,000 units (roughly 1.187 million versus 1.185 million), aided by Model Y production pauses for the refresh \[64\]. Rival tracker Focus2move scored 2024 and 2025 for the Model Y \[65\]. The defensible statement is that the Model Y and RAV4 have been the world's top two vehicles for three straight years, at roughly a million-plus units each, with the title depending on tracker methodology. That a five-year-old EV is still trading the global sales crown with the RAV4, in a year when US EV share fell to 5.8 percent \[58\], remains the single strongest evidence for the thesis that a sufficiently good EV competes with everything, not just other EVs. In Q1 2026, one of every three EVs sold in America was still a Model Y \[57\]. ### **The philosophy that produced it** The reason the Tesla-versus-Rivian capital comparison is even interesting is that Tesla's efficiency was not luck; it was a philosophy, and the question is whether it transfers. Tesla treats the factory as the product. Musk's phrase is "the machine that builds the machine," and its concrete artifacts are documented in independent engineering teardowns: Munro & Associates found that the Texas Model Y's front and rear gigacastings eliminated 172 parts and roughly 1,600 spot welds from the body \[84\]. The structural battery pack turned the most expensive component into a load-bearing chassis element. Shanghai went from permits to production in under a year for about $2 billion \[85\]. The company iterates like SpaceX because it is staffed and run like SpaceX: build, break, fix, at a cadence traditional automotive purchasing departments cannot follow. When the Model 3 ramp nearly killed the company (Musk later said Tesla was "single-digit weeks" from death and burning roughly $100 million a week at the low point \[25\]), the response was a general-assembly line under a literal tent that built roughly 20 percent of the final week's Model 3s in Q2 2018, the same week Tesla finally crossed 5,000 cars per week \[21\]. Does the advantage persist in 2026? Partly. The execution muscle demonstrably persists: Q2 2026 was Tesla's best second quarter ever at 480,126 deliveries \[23\]. What has eroded is the product pipeline that muscle serves. Tesla's 2025 "new products" were a Model Y refresh and decontented Standard trims at $36,990 and $39,990 \[24\], while the Cybertruck stayed niche and the sub-$30,000 vehicle remains unbuilt. Tesla's ramp advantage was built ramping new products. There have been no new mass-market products to ramp. And this is precisely the asymmetry Rivian is exploiting. Scaringe's R2 commentary reads like a man who studied the 2018 tent from a safe distance: launch one vehicle, with deliberately limited build combinations, on a single shift until the slowest supplier is fixed, then add the second shift \[12\]. Rivian is four to five years behind Tesla on volume by my estimate. On institutional learning about how to ramp, the gap looks smaller, because Rivian has now done two launches, and the second one is beating its own guidance \[4\]. --- ## **7\. The China Mirror: What Verbatim Copying Plus a Battery Monopoly Produces** If you want to know what a full-commitment version of the Tesla playbook looks like without American constraints, it exists, and it is why this paper keeps qualifying "the only automaker" with "Western." China's EV champions copied the Tesla formula almost verbatim: vertical integration, software-defined vehicles, in-house ADAS, direct sales, relentless cost-down. Then they added the one advantage Tesla never had: ownership of the battery supply chain. The numbers are stark. CATL and BYD together supplied over 55 percent of the world's EV battery capacity in 2025 (39.2 and 16.4 percent respectively, per SNE Research) \[60\]. China hosts roughly 85 percent of global battery cell manufacturing capacity, refines about 65 percent of the world's lithium and 75 percent of its cobalt, and processes over 90 percent of battery-grade graphite, per the IEA \[59\]. The output of that stack: BYD sold 2.26 million pure battery-electric vehicles in 2025, taking the global BEV crown from Tesla (1.64 million, down 8.6 percent) by more than 600,000 units \[61\]\[22\]. BYD's Seagull city car has sold for roughly $8,000 to $10,000 in China, and by mid-2026 a $13,000 trim carried lidar \[62\]. And BYD's God's Eye ADAS ships at no extra cost across the lineup in three tiers, up to 600 TOPS with triple lidar at the top \[42\]. Two implications for the Rivian thesis. First, it validates the strategy: vertical integration plus in-house intelligence is not a Tesla eccentricity; it is what winning looks like in the only EV market that is actually functioning at scale. Second, it defines the ceiling of the opportunity: tariffs currently keep BYD out of the US market, which means the American market is a protected arena where exactly two companies run the winning playbook. That protection is policy, and policy changes. Rivian's window to reach scale is the duration of that protection, and nobody knows how long it is. --- ## **8\. The Demand Problem: EVs After the Subsidy** Every growth claim in this paper has to survive contact with the ugliest demand environment in the modern EV era, so here is that environment, unvarnished, from Cox Automotive's data \[55\]\[56\]\[57\]\[58\]: | Period | US EV sales | EV share | Note | | ------- | -------------------------------------- | -------- | -------------------------------------------------------------------- | | FY 2024 | \~1.30M | 8.1% | Peak subsidy era | | Q3 2025 | 438,487 (record) | \~10.5% | Buyers pulled purchases forward before the credit died Sept 30, 2025 | | Q4 2025 | \~234,000 (down 46% QoQ) | 5.8% | The hangover | | Q1 2026 | 216,399 (down 27% YoY) | 5.8% | Rivian and Lucid were among the only brands growing | | Q2 2026 | 247,226 (up 14.7% QoQ, down 20.5% YoY) | \~5.8% | Stabilization, not recovery | The $7,500 federal credit for new EVs ended September 30, 2025 under the One Big Beautiful Bill Act \[55\]. Since then the market has settled at roughly 5.8 percent share, about half its Q3 2025 peak, while hybrids grow \[58\]. Electrification is, as I have written before, a market stuck in a rut: the number of EV models keeps increasing while aggregate demand has weakened. The resistance is not irrational, and I say that as an enthusiast. Out of the many people I have worked with at Tesla, only a small fraction drive a Tesla, or any EV at all. The reasons I hear are consistent: EVs skew smaller than what larger families need, and road trips gain non-trivial time. For Bay Area skiers racing to Tahoe to maximize ski days, a charging stop can add 45 minutes to the run, and often in the wrong places, at the wrong time, behind a queue of other EVs with the same idea. Anyone who dismisses this as range anxiety propaganda has not stood in a Sacramento-corridor Supercharger line on a powder Saturday. This is the demand backdrop against which Rivian raised its 2026 guidance \[4\], and it is why the R2 read-through matters more than the absolute numbers. In a market down 20 percent year over year, the brands still growing are the ones selling a product people actively want rather than a fuel type. Cox's own commentary noted that most automakers saw EV sales down 60 to 70 percent in Q1 2026 while Rivian and Lucid grew \[57\]. Scaringe's framing, from the Q2 release: "The U.S. automotive marketplace is starved for high-quality EV choice" \[6\]. Self-serving, but the Cox data does not contradict him. The other half of the demand story is that the incumbents left. GM took a $1.6 billion EV capacity charge in October 2025 and roughly $6 billion more in Q4, cancelling the BrightDrop van along the way \[71\]. Ford cancelled its three-row electric SUV, ended F-150 Lightning production, and pushed its next-generation EV truck to 2028 \[73\]. Honda cut its electrification investment plan by roughly $21 billion and abandoned its 2030 EV mix target \[72\]. Whatever else this means, it means the companies best positioned to out-manufacture Rivian chose not to compete for exactly the customers Rivian needs. The paradox of the post-subsidy market is that it is terrible for EV volume and excellent for the competitive position of the two companies still fully committed. --- ## **9\. The Brand Vacuum: Tesla's Self-Inflicted Opening** I try to keep personal distaste out of financial analysis, but the brand data is not a vibe; it is measurable, and it is a direct input to Rivian's addressable market, because a Rivian is what a disaffected Tesla customer buys without giving up the technology experience. The measurements: a March 2025 YouGov/Yahoo survey found 67 percent of Americans would not consider buying or leasing a Tesla, with most citing Musk \[66\]. The Axios Harris Poll 100 ranked Tesla's reputation 95th of 100 US brands in May 2025 \[67\]. Brand Finance estimated Tesla's brand value fell from roughly $43 billion to $27.6 billion during 2025 \[68\]. Tesla's European registrations fell about 28 percent in 2025 while the European BEV market grew \[69\], and in its home state of California, registrations fell 11.4 percent and share slipped to 9.9 percent, a seventh consecutive quarterly decline as of mid-2025 \[70\]. The partisan realignment in the YouGov panel data (Democratic consideration falling, Republican consideration rising, per a Northeastern University analysis of BrandIndex) suggests this is not a temporary news-cycle effect but a durable re-sorting of who will buy the product \[66\]. Among people I know, the pattern is blunt: they want the EV-plus-software experience and they do not want the baggage. Until 2026, there was no second vendor for that experience at a mainstream price. R2 at $45,000 to $58,000 is that second vendor, arriving at precisely the moment the first vendor's brand is at its weakest. Timing is not everything, but it is not nothing. --- ## **10\. Ownership in an Autonomous World, and the Outdoor Activities Problem** If autonomous vehicles become the predominant form of urban transportation, the number of cars people own will fall. I believe that, and the early data supports directionally without proving magnitude. Waymo went from 250,000 paid rides per week in April 2025 \[89\] to 500,000 per week across ten-plus cities by March 2026 \[76\], targets one million weekly rides by the end of 2026 \[75\], and did most of that scaling on a fleet of only about 3,000 vehicles \[90\], meaning utilization, not fleet size, drove the growth. Later today I will take a Waymo to Chase Center, and the notable thing about that sentence in 2026 is how unremarkable it has become. The economics are heading the right direction but are not there yet. Measured pricing (Obi's analysis of 94,000 ride requests in the Bay Area, January 2026) has Waymo averaging $19.69 per ride against Uber's $17.47, a premium that shrank from 31 percent to about 13 percent in nine months \[77\]. Owning a car costs the average American $11,577 a year, or 77 cents per mile at 15,000 miles per year, per AAA's 2025 study \[78\]. The projections that robotaxis reach 25 cents per mile (ARK) or halve ride-hail costs by 2030 (McKinsey) are exactly that, projections, from parties with varying incentives, and I rate them accordingly in the bibliography \[79\]\[80\]. But the direction is not in serious dispute: electric drivetrains plus no driver plus high utilization is structurally the cheapest way to move a person a mile, and as that cost falls, the share of trips that justify owning a car falls with it. This concept compounds. What the robotaxi does not solve is what I call the **outdoor activities problem**. A Waymo cannot carry your bikes to the trailhead, your raft to the put-in, or your family, dog, and ski gear to Tahoe with a cooler wedged in the trunk, and no near-term robotaxi network will position vehicles at mountain trailheads with rack hardware installed. Until autonomy solves gear, dirt, distance, and dogs, households like mine keep a vehicle, and the vehicle they keep is the one optimized for exactly those trips. This is, not coincidentally, a description of a Rivian. Given the choice between an R1S quad-motor and a Model X, I take the R1S every time, use it on weekends, and let Waymo have my Tuesday trips to Chase Center. Follow that logic to the fleet level and you get a plausible end state for private ownership: the commodity commute trip migrates to autonomous networks, and the owned vehicle bifurcates toward capability and identity. In that world the most defensible consumer franchise in the industry is the adventure vehicle, and the most exposed franchise is the commuter appliance. Tesla's product line is commuter appliances. Rivian's is not. That, more than any spec sheet, is my long-term case for the brand. --- ## **11\. The Van Business: Amazon, a Billion Miles, and the Fleet Apprenticeship** The least glamorous part of Rivian's business may be the most strategically underrated, and I will make the unfashionable claim directly: the van program's strategic value exceeds the Model X and S franchise, low margins and all. The verified numbers: Amazon ordered 100,000 electric delivery vans in 2019 with a commitment to have them on the road globally by 2030 \[48\]. As of mid-2026 there are more than 40,000 Rivian EDVs delivering for Amazon in the US \[6\]\[48\], they delivered over one billion packages in the US in 2024 alone \[48\], and the platform crossed one billion cumulative miles, a figure Rivian confirmed alongside Q2 2026 results \[6\]\[12\]. Rivian vans make up roughly 80 percent of Amazon's global electric van fleet \[48\]. Amazon's exclusivity ended in November 2023 \[50\]; AT&T became the first non-Amazon customer, and HelloFresh the first non-Amazon van fleet customer in April 2025 \[51\]. Why this matters more than the revenue it books: **First, it is the fastest reliability laboratory in the industry.** A delivery van runs stop-start duty cycles all day, every day, racking up miles and abuse at a rate no consumer will ever match. A billion miles of telemetry from vehicles that share powertrain and electronics DNA with the consumer products means Rivian discovers powertrain defects, thermal problems, and component wear years faster than a consumer fleet would surface them, and fixes flow back into R1 and R2\. Tesla never had this; its high-mileage data came slowly, from consumer outliers. **Second, it is a paid apprenticeship in fleet operations with one of the largest fleet operators on earth.** Charging depots (Amazon has built more than 50,000 chargers across 250+ delivery stations \[48\]), uptime management, service logistics, fleet telematics: these are precisely the operational competencies a robotaxi manufacturer needs, and Rivian is learning them on Amazon's dime years before its Uber robotaxis ship. The Uber deal of March 2026 \[5\] looks less surprising when you notice Rivian was already the only EV startup running a six-figure-scale commercial fleet relationship. **Third, the market is structurally excellent.** Delivery vans are bought on cost per mile, not on brand sentiment, and an electric drivetrain wins that math on high-utilization urban routes. It is a volume business insulated from both the consumer demand malaise of Section 8 and the brand politics of Section 9. As for the science fiction version, where the vans drive themselves and robots walk packages to the door: the fleet-side half of that is a straightforward extension of technology Rivian is already building, and the last-fifty-feet half is nowhere close, which is roughly Amazon's problem to solve rather than Rivian's. That future is a decade-plus out, I am not certain I want it, and the van business does not need it to justify itself. It justifies itself on packages, miles, and the data those miles generate. --- ## **12\. The OEM Lifeline Runs Both Ways: Volkswagen and Ford** An underappreciated feature of the Rivian story is that legacy OEM capital keeps arriving, and increasingly it arrives as payment for technology rather than as venture hope. **Volkswagen** is the load-bearing example. The joint venture, Rivian and VW Group Technologies, launched November 2024 with VW committing up to $5.8 billion through 2027 \[43\]. The tranches have been landing on schedule: $1 billion via convertible note in June 2024, roughly $1.3 billion at JV closing for IP and equity, $1 billion in June 2025 when Rivian hit two consecutive gross-profit quarters, and $1 billion on April 30, 2026 after the JV's zonal architecture passed winter testing in Sweden and Arizona \[43\]\[44\]\[46\]\[47\]. That is roughly $4.3 billion received of the $5.8 billion (my arithmetic across the disclosed tranches, flagged as computed). The JV employs 1,500+ engineers, and the first VW product on Rivian's architecture, the roughly 20,000-euro ID.EVERY1 family, is targeted for 2027 \[45\]. The strategic meaning is hard to overstate: the world's second-largest automaker evaluated its own software organization against a startup's and chose to pay billions for the startup's. In the process, VW became a shareholder large enough to overtake Amazon as Rivian's biggest outside holder in 2026 (per secondary reporting of SEC filings; pin to the next proxy statement before quoting a precise percentage). Rivian's software and services segment is the visible result: $515 million of Q2 2026 revenue at a 42 percent gross margin, with 60 percent of it attributable to the JV \[6\]. In quarters where the automotive side lost money, the software side is what made consolidated gross profit positive \[2\]\[6\]. Rivian is, quietly, the only EV startup with a profitable enterprise software business bolted to its side. **Ford** is the cautionary mirror. It invested $500 million in 2019 (about $1.2 billion eventually), planned a Rivian-based Lincoln that died in April 2020, cancelled all joint development in November 2021, and sold roughly 91 million shares during 2022 for around $3 billion, ending near a 1.15 percent stake by early 2023 \[52\]\[53\]\[54\]. Note the wording: Ford substantially exited; I could not verify a documented zero-stake date. Ford booked an $8.3 billion mark-to-market gain in 2021 and a $7.4 billion reversal in 2022 on the position \[54\], which tells you more about Rivian's stock chart than about the partnership. The pattern across VW, Ford, and Amazon is the thesis of this section: the industry keeps deciding, with money, that Rivian built something it cannot build itself. Sometimes the partner stays and pays (VW, Amazon), sometimes it flinches and leaves (Ford). Either way, Rivian has repeatedly monetized its engineering without surrendering control, and that is the specific mechanism by which it can keep "helping prop up" the OEMs that invested in it: they get architecture and vans; Rivian gets the capital that funds the next platform. --- ## **13\. Risks and Honest Counterarguments** A white paper that only argues one side is marketing, so here is the strongest version of the bear case, with the same sourcing standard. **The company still loses a lot of money.** Q2 2026: $837 million net loss, negative $379 million adjusted EBITDA, negative $849 million free cash flow \[6\]. Guidance implies roughly $1.8 to $2.0 billion of negative adjusted EBITDA for 2026 \[6\]. Pro forma liquidity of $7.2 billion \[6\] covers several years at this burn only if the burn shrinks on schedule. It has been shrinking; it must continue to. **Automotive gross profit is still negative.** The 11 percent consolidated margin leans on software and services, which leans on VW JV revenue recognition, some of which is amortization of a one-time $1.96 billion IP payment \[2\]. The automotive segment lost $36 million gross in Q2 2026 \[6\]. The "positive exit rate" claim for late 2026 \[12\] is a forecast from the people most incentivized to make it. **Dilution is the quiet tax.** The VW milestones, the Uber investment, and the July 2026 offering all arrive as equity or converts. Additional paid-in capital grew from $29.9 billion to $33.3 billion in eighteen months \[1\]\[6\], before the July raise. Shareholders are buying the growth with ownership. **The autonomy claims are announced, not shipped.** Eyes-off driving was planned for 2026 \[13\] and has not shipped as of this writing. The in-house silicon ships on R2 late 2026 per company statements \[14\]. Every legacy claim in this industry, Tesla's most of all, argues for pricing autonomy roadmaps at a steep discount until features are in customer hands. **Demand at scale is unproven.** R2's early conversion data covers reservation holders, the most enthusiastic possible cohort, in the launch quarter, at the $58,000 trim \[12\]. The thesis gets tested in 2027 when the $45,000 trims meet the post-subsidy mass market at 5.8 percent EV share \[58\]. **The execution tell to watch:** Rivian ran layoffs in October 2025 (about 600 people, 4.5 percent) and again in June 2026, one week after R2 deliveries began \[87\]\[88\]. The company frames both as scaling discipline. They are equally consistent with a company managing its cash runway hard. Both readings can be true. **And the wildcard:** the entire US competitive framing of this paper exists inside a tariff wall. If Chinese EVs enter the US market on anything like their home economics, the relevant comparison stops being Rivian versus Tesla and becomes everyone versus BYD \[61\]\[62\]. --- ## **14\. Conclusion** Strip the story to its skeleton and it looks like this. It took Tesla roughly $6.8 billion in today's dollars to reach its mass-market car; it took Rivian about $29.8 billion to reach the same milestone, a real 4.4x gap that inflation only partly excuses \[20\]\[1\]. What Rivian bought with the extra capital is the actual question, and the answer turns out to be: a second vehicle platform with half the bill of materials \[11\], a 215,000-unit plant with a 300,000-unit sibling under construction \[1\]\[9\], a software business that a top-two global automaker pays for and that gross-profits at 42 percent \[6\], a commercial van fleet with a billion miles of telemetry and the industry's best fleet apprenticeship \[48\]\[6\], a signed robotaxi launch customer \[5\], and the only Western-automaker autonomy program other than Tesla's that owns everything from the camera to the chip to the model \[14\]\[33\]. Tesla still holds the advantages that matter most in manufacturing: scale, ramp experience, and self-funding profitability. My estimate says Rivian is four to five years behind on volume, and nothing in this paper closes that gap by argument; only factories close it. But Tesla's product pipeline has gone quiet at exactly the moment its brand became a liability with half the buying public \[66\]\[67\], the legacy incumbents have retreated from the field \[71\]\[72\]\[73\], and the Chinese companies running the same playbook better are locked out by policy. That leaves one company positioned to absorb the demand Tesla sheds, selling the category of vehicle (adventure-capable, gear-swallowing, dog-approved) that autonomous ride networks will be last to displace. I would not call the outcome likely. I would call it live, which for a challenge to Tesla is a first. Nobody else in the West has gotten this far, this credibly, and the R2 launch quarter, beating guidance in the worst EV demand environment in a decade \[4\]\[58\], is the first hard evidence that the thesis survives contact with the market. The next two data points that matter: automotive gross profit at year-end 2026, and whether eyes-off ships. Watch those two lines, and ignore almost everything else. ## ### The Distributed Home Data Center Fallacy URL: https://nicholasjohnson.blog/the-distributed-home-data-center-fallacy/ Last updated: 2026-07-17T16:00:05.000Z **A First-Principles Engineering Critique of the SPAN / NVIDIA / PulteGroup XFRA Architecture** **Executive Summary** In April 2026, [SPAN](https://www.span.io/blog/span-announces-xfra-a-distributed-data-center-solution-to-close-the-speed-to-power-gap-for-ai-compute-demand?ref=nicholasjohnson.blog) announced [XFRA](https://www.xfra.ai/?ref=nicholasjohnson.blog), a distributed data-center architecture in which exterior-mounted compute nodes - each containing 16 NVIDIA RTX Pro 6000 Blackwell Server Edition GPUs, 4 AMD EPYC CPUs, and 3 TB of RAM - would be installed on newly built homes to harvest "unused" residential electrical headroom. The company claims it can deploy compute at \~$3M/MW in six months, versus \~$15M/MW and 3–5 years for a 100-MW centralized facility, and it intends to reach 1 GW of annual capacity by 2027 \[1\]\[2\]. This paper argues that the claim is an arithmetic fiction sustained by externalized costs. Specifically, we show: 1. **Grid layer.** The "40% spare capacity" SPAN markets as unused power - it is the engineered diversity margin that prevents pad-mount distribution transformers from cooking themselves. A 100-home pilot at SPAN's own stated 12.5 kW/node load adds 250–314 % nameplate overload to typical 25 kVA residential transformers. Scaled to 80,000 nodes, the avoided distribution upgrade liability is on the order of $0.8–1.6 B, externalized to ratepayers. 2. **Physical security layer.** A $200K rack of GPUs bolted to the exterior wall of a single-family residence cannot host any workload subject to PCI-DSS, HIPAA, SOC 2 Type II, ITAR, FedRAMP, or most enterprise contractual security postures. This is not a marketing gap; it is a category exclusion. 3. **Maintenance layer.** An industry-standard GPU AFR of \~9%, combined with the rest of the BOM, yields \~2 service events/node/year. At 80,000 nodes, that is \~160,000 fully-loaded truck rolls/year at $1,000–1,500/event - roughly $192M/year, or **2–4× the per-MW maintenance burden of a hyperscale facility** \- before accounting for the access-rights problem that single-family homes pose. 4. **Network and economics layer.** At full load, an XFRA node generates \~50–100 Mbps of sustained upstream traffic on the homeowner's residential broadband - the inverse of how PON last-mile is provisioned. Energy delivered to compute costs \~$0.36/kWh at California residential rates, versus $0.05–0.08/kWh at industrial-tariff hyperscale rates. The proposition is **5–7× more expensive on energy alone**, before maintenance, capacity-factor losses from load-shedding, and security-tier workload exclusions. The XFRA proposal does not survive first-principles engineering. It survives only as a marketing wrapper around SPAN's existing - and genuinely useful - smart-panel and new-construction installation channel. We characterize this as *innovation-themed* infrastructure: a narrative engineered to access A.I. capital markets rather than to deliver compute economics. --- ## **1\. Grid Impact: The Diversity-Factor Fallacy** ### **1.1 What SPAN is actually claiming** Per SPAN's own statements, each XFRA node is a Dell PowerEdge server with 16× RTX Pro 6000 Blackwell Server Edition GPUs at 450–600W each, 4× EPYC CPUs, 3 TB of DDR5 memory, networking, PSU losses, and an inline 16 kWh battery. SPAN's own pilot disclosure to Latitude Media specifies **1.25 MW across 100 nodes = 12.5 kW/node, continuous** \[2\]. This is sized to the 600W boost-mode GPU TDP, not the 250W typical-inference draw - i.e., the marketing math assumes peak utilization 24/7. SPAN's foundational claim is that "the average home operates at 40% of nameplate capacity, leaving \~80 A (≈19.2 kW at 240V) of headroom that can be harvested" \[3\]. This claim conflates two distinct quantities: **individual peak capacity** (the panel rating) and **system diversified capacity** (what the distribution transformer is sized for). ### **1.2 The actual distribution engineering** A typical American suburban pad-mount distribution transformer is 25–50 kVA, serving 4–8 single-family homes. It is sized using a **diversity factor** of 0.4–0.6, reflecting the empirical reality that homes do not run their dryers, ovens, HVAC compressors, and EV chargers simultaneously, and that across multiple homes the individual peaks de-correlate. The arithmetic, for a representative 25 kVA transformer serving 5 homes with 200 A (48 kW) service: | Quantity | Value | | ------------------------------ | ----------------------------------------------- | | Per-home individual peak | \~8 kW (coincident demand, summer afternoon) | | Diversity factor (5 homes) | \~0.4 | | Coincident peak at transformer | 5 × 8 × 0.4 = **16 kW** | | Transformer nameplate | **25 kVA** | | Headroom (engineered margin) | 9 kVA (36% - the "underutilization" SPAN cites) | Now overlay 5 XFRA nodes: | Quantity | Value | | ---------------------------------- | --------------------- | | XFRA continuous load (5 × 12.5 kW) | **62.5 kW** | | Diversified residential load | 16 kW | | **Total continuous demand** | **78.5 kW** | | Transformer nameplate | 25 kVA | | **Loading factor** | **314% of nameplate** | The transformer will run continuously at over three times its design rating. Per IEEE C57.91 thermal modeling, every 8 °C above design winding temperature halves the insulation life. A 25 kVA transformer at 314% load will hit thermal runaway in days, not years. Even isolating just the XFRA load (62.5 kW on 25 kVA = 250 %), there is no operating regime in which this is survivable without **distribution upgrades;** something the US has already been doing so "well". ### **1.3 Aggregated impact on PG&E** PG&E's own published data shows the system is **45% utilized on average** \[4\]. PG&E uses this same figure to argue that *centralized* data-center load can be absorbed efficiently - exactly the opposite use of the statistic SPAN is making. The 55% headroom is what allows: - N-1 contingency planning (any single feeder can fail without cascading) - Coincident summer peak absorption (47°C interior temperatures during heat domes) - Wildfire PSPS reconfiguration - EV adoption ramp without rebuilds CAISO forecasts peak demand growing from **46.094 GW in 2025 to 52.94 GW by 2030** \- a 15% increase requiring \~$30B in transmission and distribution upgrades \[4\]\[5\]. PG&E's data-center pipeline alone is **10 GW** \[5\], and PG&E has announced a **$73B grid infrastructure upgrade plan** to accommodate it \[4\]. Against this backdrop, SPAN's 1 GW-by-2027 target - added as **uncoordinated continuous residential baseload** \- is not "free use of underutilized capacity." It is a parallel grid buildout obligation, with the bill landing in the utility's distribution engineering budget. ### **1.4 The 100-home pilot quantified** For a single 100-home XFRA pilot in a southwestern subdivision: - 100 nodes ÷ 5 homes/transformer = **20 transformers requiring upgrade** - Upgrade cost (25 → 100 kVA pad-mount + secondary cable + outage labor): \~$50K–100K per transformer - **Pilot-level distribution upgrade cost: $1M–2M** (not in SPAN's $3M/MW figure) - Per-node externalized cost: **$10,000–20,000**, on top of the \~$3,750/node SPAN cites - Plus: feeder reconductoring, substation transformer capacity studies, protection coordination Scaled to 80,000 nodes: **$800M–$1.6 B in externalized distribution upgrades.** The $3M/MW vs. $15M/MW claim becomes $6–8M/MW vs. $15M/MW once internalized, and the gap collapses further once load-shedding capacity-factor losses are priced in (Section 4). ### **1.5 The most damning detail** In March 2025, PG&E launched the SAVE virtual power plant program **with SPAN as an aggregator**, using SPAN smart panels to **shed residential load** during peak \[6\]. The exact same hardware platform that SPAN now markets to add 12.5 kW of always-on load to a home was, twelve months earlier, marketed to remove load from the same home during the same hours. Both pitches cannot be simultaneously true. The grid does not have the capacity SPAN claims; SPAN's own contract with PG&E presupposes this. --- ## **2\. Physical Security: A Category Exclusion** Data centers exist as fortresses for reasons rooted in silicon, not in aesthetics. The XFRA architecture eliminates every layer of that fortress. ### **2.1 What a Tier III/IV data center provides** - Perimeter fence, anti-ram barriers, vehicle inspection - 24/7 manned security with badge + biometric + mantrap entry - Per-rack locking cages with audit logs - CCTV with N-year retention on every cabinet - No personal electronics; no removable media; no unsupervised access - Network ports physically restricted; out-of-band management on isolated VLAN - Tamper-evident seals on every chassis - TPM 2.0 + measured boot + remote attestation chain These controls exist because at the silicon and firmware level, **physical access defeats most cryptographic protections**: | Attack class | What it requires | | ---------------------------------------- | ------------------------------------- | | Cold-boot DRAM key extraction | <60s of physical access post-shutdown | | Spectre/Meltdown/Rowhammer side-channels | Local code execution + time | | Voltage / clock glitching | Hardware tap on PSU rails | | PCIe DMA attack | A free PCIe/Thunderbolt slot | | JTAG debug interface | Physical port + low-level tools | | Network tap / fiber splicing | Access to cable run | | Firmware re-flash / supply-chain swap | Chassis access + reboot window | ### **2.2 What an XFRA node provides** A box on the exterior wall of a single-family residence, accessible from the yard. The homeowner has unrestricted approach. The driveway has unrestricted access. There is no cage, no mantrap, no audit log, no 24/7 surveillance, and no plausible mechanism to enforce any of those. Each node contains roughly $144K in GPU silicon alone (16 × $9,050 retail MSRP for the Pro 6000) - an attractive target for theft independent of any data-extraction motive. ### **2.3 The workload exclusion** A non-exhaustive list of workloads that **cannot legally or contractually** run on hardware in this security posture: - **PCI-DSS** (any payment-processing inference, including fraud detection) - **HIPAA** (any healthcare inference, including imaging and transcription) - **ITAR / EAR** (any defense, aerospace, or controlled-technology workload) - **SOC 2 Type II** (most enterprise SaaS customers require this of their vendors) - **FedRAMP** (any US government workload) - **GDPR Article 32** (any EU personal-data processing - requires "appropriate technical and organizational measures") - **Sovereign A.I.** (the entire category of nation-state-grade workloads, ironically what NVIDIA's Marc Spieler markets elsewhere) What remains: low-stakes consumer chatbot inference and cloud gaming. This is precisely the workload tier most aggressively commoditized on a $/token basis, where margins are already negative for non-vertically integrated providers. It is the worst possible TAM to attack with a 5–7× cost-disadvantaged platform. The argument that "the data is encrypted in flight and at rest" is non-responsive. Confidential computing requires a hardware root of trust (Intel TDX, AMD SEV-SNP, NVIDIA confidential compute mode) **and** physical assurance that the silicon itself has not been swapped, glitched, or tapped. Physical assurance is the part the home environment cannot provide. --- ## **3\. Maintenance: The Truck-Roll Apocalypse** ### **3.1 Failure rate baseline** Field data from Meta's H100 training cluster (16,384 GPUs, 54-day window) showed 466 unplanned interruptions, of which \~78% were hardware-related, with GPUs the most common single failure mode \[7\]. The industry consensus translates this to a **\~9% annual GPU failure rate** under high-utilization production load, with cumulative three-year risk exceeding 25% \[8\]. Independent vendor data place the average GPU MTBF at approximately 20,000 hours \[9\], corresponding to a similar AFR. For a single XFRA node with 16 GPUs: - GPU-only failures expected: 16 × 0.09 = **1.44 events/year** - CPU failures (4 × 3% AFR): 0.12 events/year - DIMM failures (typical 1–2%/yr × \~24 modules): \~0.4 events/year - PSU / fan / cold-plate pump/battery / smart-panel: \~0.3 events/year - **Total expected service events per node per year: \~2.0–2.3** ### **3.2 Truck-roll economics** Field-service industry data place the direct truck-roll cost at $200–500 per dispatch, with the fully loaded cost (labor, vehicle, dispatch overhead, opportunity cost) approaching $1,000/dispatch \[10\]\[11\]. For data-center-grade work requiring a qualified hardware technician with parts inventory, conservatively: | Component | Cost | | --------------------------------------------------- | ------------ | | Skilled DC technician (4 hr × $150/hr fully loaded) | $600 | | Truck, tools, parts kit | $200 | | Dispatch, scheduling, coordination | $100 | | Expected return-visit overhead (25% × $1,000) | $250 | | Spare parts | $200 | | **Loaded cost per event** | **\~$1,350** | ### **3.3 Scaling** | Metric | 100-home pilot | 80,000-node target | | ------------------------- | -------------- | ------------------ | | Service events/year | 200–230 | 160,000–184,000 | | Annual truck-roll cost | $270K–$310K | **$216M–$248M** | | Per-MW maintenance burden | \~$216K/MW/yr | \~$216K/MW/yr | A reference hyperscale 100 MW data center runs total maintenance (including labor, parts, and facility upkeep) on the order of **$50K–$100K/MW/year**. XFRA's truck-roll-only maintenance cost is **2–4× the** hyperscale alternative's entire maintenance budget. ### **3.4 The access-rights problem** The cost number above assumes the technician can simply arrive and work. In reality, a single-family-home installation creates a *legal and operational constraint* that data centers do not have: - Easements covering exterior-mounted equipment do not grant 24/7 interior access. - Homeowners travel, sleep, work from home, have children, have pets, are sick, have guests, have hostile family members. - A 2 AM dispatch to fix a failed GPU is socially impossible, even if it is technically necessary. - Liability for property damage during service is uninsurable at scale (one technician slipping in one homeowner's bathroom creates a class-action exposure). - Background-check, bonding, and licensing requirements vary by jurisdiction. - A homeowner who is angry, intoxicated, traveling, or simply unresponsive can block a service event indefinitely. The practical consequence is that **MTTR (mean time to repair) balloons from 2–4 hours (data center) to 24–96 hours (home)**, with corresponding reductions in capacity factor and SLA performance. Hyperscale inference customers pay for p99 latency at five 9s of availability. That SLA is not defensible when the dispatch window depends on whether the host family is on vacation. --- ## **4\. Network, Energy, and Load-Profile Economics** ### **4.1 The bandwidth assumption is inverted** LLM inference economics target 40–60 tokens/sec for real-time UX \[12\]. Per Pro 6000, one can plausibly serve \~50–100 concurrent inference sessions on memory-bound workloads. For 16 GPUs/node: | Quantity | Value | | ------------------------------------------- | ----------------------------------------------------- | | concurrent Sessions/node | 800-1600 | | Bytes/Token (Typical) | \~4 | | Per Session Sustained Bandwidth | 5-50 KB/s (incl. KV-Cache, prmpt context, telemetery) | | Addregated per node sustained upstream | 50-100 Mbps continous | | Plus model swaps, prefix caches, monitoring | +20-30 Mpbps | Now compare this to how residential broadband is actually built. Last-mile **PON architecture with a 32:1 or 64:1 split ratio:** one fiber strand carries 32–64 customers \[13\]. Median US residential connection is roughly **250 Mbps down / 25 Mbps up** \[14\] - asymmetric because the design assumption is that homes are net consumers, not producers, of bandwidth. XFRA inverts this assumption. Each node is a server generating sustained upstream traffic that exceeds the upload caps typical for residential plans. Five XFRA-equipped homes on a single PON segment will continuously saturate the segment's upstream. ISPs will respond by: 1. Capping the customer (kills XFRA), 2. Billing XFRA upstream traffic on commercial tariffs (kills the economics), or 3. Requiring middle-mile upgrades that SPAN does not budget for. The Corning **long-haul backbones are already approaching Shannon's limit** growth \[15\]. XFRA loads the *same* backbones via residential ISPs that are not contracted to carry server-grade upstream. ### **4.2 The energy-cost catastrophe** Compare cost-per-kWh delivered to compute: | Operator class | Generation rate | Delivery / total | PUE | $/kWh to compute | | ---------------------------------- | --------------- | ---------------- | --------- | ---------------- | | Hyperscale (Tier 3 industrial) | $0.04–0.06 | $0.05–0.08 | 1.15–1.20 | **$0.06–$0.10** | | CA Commercial B-1 (PG&E) | $0.12 | $0.41 | 1.10–1.15 | $0.45 | | **CA Residential (PG&E, average)** | mixed | **$0.33** | 1.10 | **$0.36** | | CA Residential **TOU peak 4–9 PM** | \- | **$0.50+** | 1.10 | **$0.55** | \*Sources: \[16\]\[17\]\[18\]\[19\] XFRA energy is **5–7× more expensive than hyperscale energy on average**, and **6–9× more expensive during the exact 4–9 PM window when consumer inference demand peaks**. SPAN claims to compensate homeowners with discounted electricity or a $150/month payment - but the meter is the homeowner's, the rate is residential, and the unit economics of inference do not have room to absorb a 6× energy multiplier. ### **4.3 The load-profile death spiral** Hyperscale data centers operate at high capacity factor (≥90%) because they own their utility interconnection and pay a flat industrial tariff. Residential service does not allow this: - Residential coincident peak: **4–9 PM weekdays**. - Consumer inference demand peak: **4–9 PM weekdays** (literally the same hours). - To avoid overloading the home panel and dodge TOU peak rates, XFRA *must* shed compute during the hours when token revenue is highest. - Shedding 5 hours/day × \~250 weekday-peak days ≈ 1,250 hr/year = **14% capacity factor loss** from TOU alone. - Add residential vacation periods, HVAC compressor-startup overlaps, EV charging conflicts, and weekend daytime peaks: realistic capacity factor lands at **65–75%**, vs. 90–95% for hyperscale. The $3M/MW capex claim assumes a denominator of *nameplate* capacity. The *available* capacity is 25–30% lower. Adjusted true capex: **$4M–$4.5M/MW**, before distribution upgrades and maintenance overhead are internalized. ### **4.4 The integrated comparison** | Metric | Hyperscale 100 MW | XFRA 80,000 nodes (1 GW) | | -------------------------------- | ----------------- | ----------------------------- | | Nominal capex | $1.5B ($15M/MW) | $3B ($3M/MW *claimed*) | | Distribution upgrade externality | included | +$0.8B–$1.6B | | **True capex** | $1.5B | $3.8B–$4.6B | | Energy cost $/kWh to compute | $0.06–$0.10 | $0.36–$0.55 | | Maintenance $/MW/yr | $50K–$100K | $200K+ | | Capacity factor | 90–95% | 65–75% | | Addressable workloads | All tiers | Consumer-grade only | | Permitting / siting timeline | 3–5 yr | 6 mo (the one real advantage) | The only **XFRA** advantage that survives scrutiny is **deployment speed**. That advantage is real and valuable. But it is not an advantage in compute economics - it is an advantage in *capital-deployment optionality*. SPAN can move capital into operation in 6 months while a hyperscaler is still in EIS review. That has value to a financier; **it does not have value to a workload.** --- ## **5\. Revenue from XFRA Claimed** To understand the rush to compute, let's look at the proposed business model and model it from first principles. The SPAN thesis is that a homeowner hosts a standard compute node, NVIDIA silicon plus an AMD EPYC host, and sells that capacity into an aggregated inference network, sharing the proceeds with SPAN and the homebuilder. The question is whether a single node clears its own operating costs. The standard unit and its market price. The reference node is an NVIDIA RTX 6000 Ada (48 GB) paired with an AMD EPYC host, the same class of silicon as the SPAN and NVIDIA reference. On the actual peer-supply compute marketplaces that a distributed network would have to clear against, this card rents for roughly $0.60/hr ([Vast.ai](http://vast.ai/?ref=nicholasjohnson.blog) median, range $0.39–$1.00) to $0.77/hr (RunPod), inclusive of the CPU host. That marketplace rate is the ceiling on what a home node can earn, because a distributed home fleet is a price taker competing directly against those same listings - it cannot charge hyperscale contract prices for consumer-grade, interruptible capacity. We use $0.60/hr as the base case. Where the revenue leaks out. A node cannot bill 8,760 hours a year. Three of the whitepaper’s own findings compound against it. First, Section 4.3 establishes that the node must shed compute during the 4–9 PM window (1,825 hr/yr) to avoid overloading the home panel and TOU peak rates — and that window is exactly when inference demand and clearing prices are highest, so the node is forced offline during its most valuable hours. Second, consumer/prosumer GPUs on peer marketplaces sit idle most of the time; realized paid utilization of 35–40% is generous. Third, the marketplace itself takes a 15–25% commission before the host sees a dollar. Net billable time lands near 3,000 hours a year at a below-peak blended rate. The cost side is worse. The node draws \~700 W (RTX 6000 Ada \~300 W + EPYC host, drives, and PSU losses \~400 W), or \~770 W at the wall once the PUE 1.10 from Section 4.2 is applied. To remain rentable, it must stay powered and reachable \~24/7- idle silicon still burns \~250–300 W. At the residential PG&E blended rate of $0.36/kWh (Section 4.2), a node that is powered continuously spends far more on electricity than it collects in rent, and any hour it runs in the 4 to 9 PM window is billed at $0.55/kWh. Layer on the maintenance/truck-roll allocation ($200K/MW/yr from Section 4.4, ≈ $154/yr for a 770 W node) and hardware amortization (\~$10,800 of silicon over a 4-year life ≈ $2,700/yr), and the unit is deeply upside-down before SPAN, or the homebuilder takes their cut. Per-unit annual economics (single RTX 6000 Ada + EPYC node, base case): | Line item (per node, per year) | Base case | | -------------------------------------------------- | ------------------- | | Marketplace rate (RTX 6000 Ada + EPYC host) | $0.60 / GPU-hr | | Theoretical max revenue (8,760 hr) | $5,256 | | Less: forced 4–9 PM shed (1,825 hr, peak-value) | −$1,095 | | Realized paid utilization (\~35%; idle otherwise) | \~3,000 billable hr | | Gross booking value | \~$1,800 | | Less: marketplace commission (\~20%) | −$360 | | Net revenue to node | ≈ $1,440 | | Energy cost (0.77 kW × 8,760 hr × $0.36, on 24/7) | −$2,428 | | Maintenance / truck-roll share ($200K/MW/yr) | −$154 | | Hardware amortization (\~$10,800 / 4-yr life) | −$2,700 | | Net cash flow per node (before SPAN/builder split) | ≈ −$3,842 | Read the last two rows together. The node collects roughly $1,440 a year and spends roughly $5,282 a year to earn about $2,428 in electricity, $154 in field maintenance, and $2,700 in hardware depreciation. It runs at roughly $3,842 per node per year on a fully loaded basis and remains negative on cash alone (revenue of \~$1,440 against \~$2,582 in cash, energy, and maintenance) even if the hardware is treated as a sunk gift. Crucially, this is before SPAN, NVIDIA, and PulteGroup take their share of the revenue and before the homeowner is paid the promised $150/month incentive, a payment which, on its own ($1,800/yr), exceeds the node’s entire net revenue. Extrapolated to the fleet. The white paper’s headline is 80,000 nodes totaling 1 GW. Scaling the single-unit result across that fleet turns a bad unit into a structurally impossible business: because the loss is per-unit and roughly linear, each additional node widens the operating deficit rather than diluting it. There is no volume at which the model crosses into positive cash flow; scale simply multiplies the hole. The table below extrapolates the base-case unit economics (expressed per GPU-equivalent) to the full 80,000-unit fleet. | Metric | Per node/yr | Fleet: 80,000 nodes/yr | | ---------------------------------------------- | ----------- | ---------------------- | | Net revenue (post-commission) | $1,440 | $115M | | Energy + maintenance (cash opex) | −$2,582 | −$207M | | Cash flow (pre-capex, pre-splits) | −$1,142 | −$91M | | Fully loaded net (incl. Hardware amortization) | −$3,842 | −$307M | | Homeowner incentives ($150/mo) | −$1,800 | −$144M | The signs are all negative, and they stay negative. On cash operating terms alone the fleet loses roughly $91M/year; once the promised homeowner incentives are added the operating drain approaches $235M/year, and on a fully-loaded basis (including the \~$864M of GPU/CPU capex depreciated over four years) the annual loss is on the order of $307M before SPAN, NVIDIA, or PulteGroup extract any margin. This is the revenue answer to the “rush to compute”: at residential energy prices, with peak-hour load-shedding baked in, and against a marketplace clearing price the fleet cannot exceed, the distributed-home compute business runs in structural negative cash flow at every scale. The model only makes sense as a capital-formation-and-channel story - not as a compute-economics story. (Note: SPAN’s own reference node bundles \~16 GPUs and 4 EPYCs at \~12.5 kW; the per-GPU-equivalent economics above scale to that node and to the 80,000-node fleet without changing sign.) ## **6\. Conclusion: Innovation-Themed Infrastructure** Three things are simultaneously true: 1. **SPAN's underlying business is real.** The smart-panel platform, the homebuilder partnership channel with PulteGroup, and the install-during-construction logistics are a genuine and defensible distribution moat. Bolting computers onto that channel uses real infrastructure. 2. **The siting problem is real.** With 14 US states considering data-center moratoria and 47% of Americans opposed to hyperscale facilities in their neighborhoods \[20\], the political path for centralized siting is constrained, creating genuine demand for alternative form factors. 3. **The XFRA technical claim is not real.** The economics break at the distribution-transformer layer, the security layer, the maintenance layer, the network layer, and the energy-rate layer - independently and severally. Any one of these would be disqualifying. All five are disqualified simultaneously. The narrative purpose of the XFRA announcement is to map SPAN's existing channel onto the A.I. capital-formation cycle. The 1 GW-by-2027 target is the kind of number that produces a Series D markup; it is not the kind of number that produces 80,000 functioning compute nodes. Telco-edge operators, who already own distributed power, fiber, security, and field-service organizations, will absorb the actual middle-tier inference market that SPAN claims \[21\]. SPAN's optimal outcome is acquisition by a hyperscaler for the smart-panel channel, with the XFRA pitch quietly retired. The diagnosis is harsher than "this won't work." The diagnosis is that the proposal does not engage with first-principles distribution engineering, hardware reliability, physical security, or residential broadband economics. While I'm a fan of smart power management, I think too many startups right now are clamoring for A.I. infrastructure funding without fully understanding the economics of GPUs. --- ## **References** \[1\] SPAN, "XFRA Distributed Data Center Solution," BusinessWire press release, April 14, 2026\. \[2\] Latitude Media, "Span to launch distributed AI data centers for edge compute," April 14, 2026\. \[3\] PV Magazine USA, "Span and Nvidia to develop AI data centers in your backyard," April 15, 2026\. \[4\] PG&E Corporate Sustainability Report 2025; PG&E SAVE/Rule 30 filings. \[5\] Data Center Dynamics, "PG&E announces $73bn grid infrastructure upgrade plan," May 2026\. \[6\] PG&E / Stocktitan, "PG&E Launches SAVE Virtual Power Plant Program," March 24, 2025\. \[7\] Meta, "The Llama 3 Herd of Models" technical report, July 2024\. \[8\] Power Policy, "The Puzzle of Low Data Center Utilization Rates," August 2025\. \[9\] Data Center Knowledge, "Scaling the Memory Wall: HBM, CXL, and the New GPU Playbook," May 2026\. \[10\] Technology & Services Industry Association, truck-roll cost analysis (cited in SightCall, CareAR field-service research). \[11\] Blitzz / Apizee field-service truck-roll cost analysis, 2026\. \[12\] 650 Group, "Interconnect Needs for LLM Inference," 2024\. \[13\] Reid Consulting Group, "Broadband 101," July 2025\. \[14\] FCC Broadband Data Collection, mid-2025 release. \[15\] Corning, "Broadband Industry Trends and Future Predictions 2025." \[16\] EnergyBot, California electricity rates summary, May 2026\. \[17\] PG&E – MCE Joint Rate Comparison, March 2026\. \[18\] Yale Clean Energy Forum, "How Hyperscalers Are Powering Their Data Centers," November 2025\. \[19\] CPUC residential and commercial rate disclosures, 2026\. \[20\] A.I. Consulting Network, "Home AI Data Centers: PulteGroup Nvidia Span CRE Impact," May 2026\. \[21\] Network World, "Startup SPAN teams with Nvidia to put data center nodes in your backyard," May 2026 - citing analyst Cordovil on the telco-edge alternative. ### Could extended-range EVs encourage more car buyers to opt for full electric? URL: https://nicholasjohnson.blog/could-extended-range-evs-encourage-more-car-buyers-to-opt-for-full-electric/ Last updated: 2026-04-19T02:04:32.000Z **Problem:** The fundamental idea here is that current EVs with 300-400-mile ranges still don't work for large SUVs that Americans actually buy. No actual pickup truck has been built that's comparable to an F150 or F250 capable of doing real work in the United States. This has left most electric vehicles in the small-to-medium car size range in America. While other markets, such as Europe and China, are content with smaller cars, the US's car size has grown significantly, making it challenging for EVs to establish a foothold in the large car segment due to efficiency issues and limited battery capacity. Today, the status quo electric vehicle typically has a range of about 300-400 miles. The added cost is requiring either 100-150kWh of the pack to move an electric vehicle down the road for that number of miles. The true irony here is that most trips are less than 40 miles every single day, which means that the majority of actual EV owners are only using a fraction of their battery capacity every day and lugging around the weight and materials. This leads to extreme overuse of precious metals and materials locked inside batteries in thousands to millions of electric vehicles, which are used only a fraction of the time (maybe for long road trips or long days on the road). The proposed solution here is to figure out a way to reduce the battery size so that the same number of batteries and the current limitations of battery capacity and manufacturing can build more vehicles, while eliminating range anxiety and the need for road-trippers to use DC fast charging more often, relying more on low-power Level 2 at-home charging at scale. **A paradigm shift can be quickly thought out. Let's actually work through the math and figure out what we're talking about.** What is the overall cost of DC fast charging in the United States? How much money have Tesla and other EV companies like Electrify America and EVgo spent on deploying an EV fast charging network? Could that money have been spent to install more at-home Level 2 chargers for more drivers across every single demographic? This includes multi-family, workplace, and home. Next, let's look at the vehicle's overall cost. A majority of the cost of electric vehicles today is actually in the battery itself. All other cars have similar materials (doors, steering wheels, radios, technology, etc.). These things are fringe benefits of a vehicle and are relatively consistent across all vehicles. The primary difference between an electric vehicle and an internal combustion engine (ICE) vehicle lies in the powertrain. In a gas vehicle, the fuel is not purchased at the time of purchase, but is purchased throughout the vehicle's lifespan (filling up every few weeks). The challenge here is that the energy density is significantly higher than what battery packs can store. To make this matter even more challenging, we aim to be a system that operates entirely on electricity. Meaning that for the majority of drive-hands, it's pulling energy from a battery or supercapacitor that is being recharged, not by an internal combustion engine but by some sort of generator. Think of a natural gas turbine or a petrol turbine that burns a little bit of fuel to generate electricity that then charges the battery at a steady state. We'll have to calculate the vehicle's steady-state motion at its extremes. This is usually around 70 to 80 miles per hour, where wind resistance becomes a significant factor in the car's speed and the amount of energy needed to keep it in motion. The other factors that are consistent across all vehicles are aerodynamics, rolling resistance, and tire-road friction. Otherwise, all vehicles are relatively the same, and therefore, the energy required to move them is a straightforward application of physics. ![](https://nicholasjohnson.blog/content/images/2026/02/01_D2-Traveler-Exterior-Major-4974.jpg) Teh 2026 Scott is offered with a Range Extender to get 500 miles of driverable range between fill ups. The challenge here is that the cost needs to be significantly lower than that of a battery-electric vehicle, and we don't believe battery costs will decrease to that level. Third, the actual user experience needs to be better, meaning that the overall requirements for a user are as close to the experience of an electric vehicle as possible, while reducing the need for DC fast charging every 2-3 hundred miles while road-tripping, and reducing the amount of time people spend getting from point A to point B. These two factors alone could significantly impact a user's experience with these vehicles. We would then assess the impacts of towing and determine whether we need a larger or smaller generator to meet our objectives. ![](https://nicholasjohnson.blog/content/images/2026/02/data-src-image-67423a16-c366-4322-898b-2984f62f347c.png) Zhejiang Geely Holding Group has launched its sustainable-experience architecture (SEA), which it claims is the world’s first open-source electric-vehicle (EV) platform. SEA will be deployed across the manufacturing group’s nine global automotive brands, beginning with Lynk & Co. ### Core model (steady speed, level road) **Forces opposing motion:** - Aerodynamic drag: $F\_d = \\frac{1}{2}\\rho C\_d A v^2$ - Rolling resistance: $F\_r = C\_{rr} m g$ **Wheel power required:** $$P\_{\\text{wheel}} = v(F\_d + F\_r) = v\\left(\\frac{1}{2}\\rho C\_d A v^2 + C\_{rr}mg\\right)$$ **Energy per distance (at the wheels):** $$\\frac{E}{\\text{distance}} = F\_d + F\_r \\quad \\Rightarrow \\quad \\text{Wh/mi}\_{\\text{wheels}} = \\frac{(F\_d + F\_r)\\times 1609}{3600}$$ **From wheels to "fuel" or battery:** $$\\text{Wh/mi}\_{\\text{battery}} = \\frac{\\text{Wh/mi}\_{\\text{wheels}}}{\\eta\_{\\text{BEV}}}, \\quad \\text{Wh/mi}\_{\\text{fuel}} = \\frac{\\text{Wh/mi}\_{\\text{wheels}}}{\\eta\_{\\text{ICE}}}$$ $$\\text{MPG} = \\frac{33700\\,\\text{Wh/gal}}{\\text{Wh/mi}\_{\\text{fuel}}}$$ Based on this data, it's clear that several factors need to be considered. Once we understand the total watt-hours or kilowatt-hours per mile equivalent to MPG, we can understand both fuel and electric vehicles' energy needs. Regardless if the energy is coming from gasoline or electricity, the efficiency of electric motors is higher. What we find is that about 2-400 watt-hours per mile is required. Therefore, we can work out 60-70 and 80 mile-an-hour averages that would need to be generated for every hour the car is driving. This means a generator would need to be in the 25-30 kilowatt range to generate the energy needed to keep a car at an optimal steady state without losing charge, say, at 70-80 miles per hour for long stretches of highway, assuming continuous generation during those times. The rest of the time, the car could either be recharging or recharging. | kWh/mile @ 60 mph | kWh/mile @ 70 mph | kWh/mile @ 80 mph | | ----------------- | ----------------- | ----------------- | | 22.9 | 26.8 | 30.6 | **Average Energy used for EVs:** | EV Make and Model | Avg Wh/mi | | ----------------- | --------- | | Tesla Model S | 320 | | Tesla Model 3 | 260 | | Tesla Model X | 340 | | Tesla Model Y | 280 | | F150 lightning | 460 | | Nissan Leaf 2023 | 320 | | Chevy Bolt | 290 | | Hyundai Kona | 262 | | Kia EV6 | 296 | | Rivian R1S | 490 | | Rivian R1T | 520 | | GM Hummer EV | 750 | Average Energy Used for ICEs: - Rav4 - CRV - Expemidition ### Concrete example (typical midsize car) Assume: $m = 1800\\,\\text{kg}$, $C\_d A = 0.65\\,\\text{m}^2$, $C\_{rr} = 0.010$, $\\rho = 1.225\\,\\text{kg/m}^3$. - **70 mph** → wheel power ≈ 17.7 kW; wheel energy ≈ 253 Wh/mi. BEV (≈90% eff.): \~281 Wh/mi from battery. ICE (≈25% eff.): \~1013 Wh/mi fuel → \~33 mpg. - **75 mph** → 20.9 kW; 279 Wh/mi wheels → \~310 Wh/mi BEV, \~1116 Wh/mi ICE → \~30 mpg. - **80 mph** → 24.5 kW; 307 Wh/mi wheels → \~341 Wh/mi BEV, \~1226 Wh/mi ICE → \~27.5 mpg. To calculate the energy consumption of a Ford Expedition in Wh/mile, you must first convert its fuel efficiency (in miles per gallon) to kilowatt-hours per gallon, then divide by 1,000 to get the watt-hours per mile (Wh/mile). This calculation is based on the vehicle's documented fuel consumption, rather than its physical characteristics. **Assumptions for Calculation** Fuel efficiency: A 2024 rear-wheel-drive (RWD) Ford Expedition averages 19 miles per gallon (combined city/highway). Energy content of gasoline: One gallon of gasoline contains approximately 33.7 kWh of energy. Engine efficiency: Gasoline engines are not 100% efficient. The average modern internal combustion engine has an efficiency of about 25%. For this calculation, we will use this value to represent the engine's conversion of gasoline energy to power. Calculation of Wh/mile **Calculate the energy consumed per mile in kWh.** Divide the energy content of one gallon of gasoline (33.7 kWh) by the vehicle's miles per gallon (19 MPG). 33.7 kWh / 19 miles = 1.77 kWh/mile **Adjust for the engine's efficiency.** Divide the energy consumed per mile by the engine's efficiency (25%). This will give you the actual power the vehicle is using to move. 1.77 kWh/mile \* 25% = 0.44 kWh/mile **Convert to Wh/mile.** Multiply the result in kWh/mile by 1,000 to convert to Wh/mile. 0.44 kWh/mile \* 1,000 Wh/kWh = 440 Wh/mile A key market segment that might embrace EREVs is EV owners who are considering switching back to an ICE due to frustration with inadequate charging availability and limited driving range in their current vehicles. In the 2024 McKinsey Mobility Consumer Pulse Survey, for example, 46 percent of US EV owners and 19 percent of European EV owners reported they were considering switching back to an ICE vehicle.6 Despite EREVs’ apparent appeal to a variety of car buyers, consumer education that clearly conveys the benefits of EREVs and generally demystifies the distinctions between all EV and hybrid-vehicle options is vital. Consumers have difficulty understanding how EREVs differ from PHEVs, BEVs, or other hybrid vehicles. Consumers in the United States appear to find the distinctions between different EV and hybrid powertrains especially perplexing. Among US car buyers included in McKinsey’s survey sample, nearly half (48 percent) agreed with the statement “I’m overwhelmed by the number of powertrains (currently available) to choose from.”7 Currently, there are few EREVs in the global market. In the United States, EREVs in the SUV and truck segment have been announced, including the 2025 Ram 1500 Ramcharger, which reports a 145-mile pure electric and a 690-mile total driving range.8 In China, Li Auto has introduced several EREVs, including its L9, which reports a 134-mile electric range and an 817-mile total range.9 And AITO’s M9 reports a 140- to 170-mile electric and 840- to 871-mile total range.10 VW-backed Scout Motors has also announced several EREV models that, according to the company, have received considerably more deposits than their Terra and Traveler BEVs.11 An electric range of 100–200 miles would meet most drivers’ daily commuting needs, while a total range of 350–600 miles could eliminate range anxiety. This may indicate a sweet spot for the EREV market (Exhibit 2). ![](https://nicholasjohnson.blog/content/images/2026/02/data-src-image-34ad5785-1b29-4825-b3d0-82cbda01d8d1.png) ![](https://nicholasjohnson.blog/content/images/2026/02/data-src-image-ca791992-e5fe-4701-9fe0-b230ed0711a3.png) The zero-emissions vehicle deadline under current EU regulations means EREVs could be sold in the region through 2034\. OEMs will need to consider their narrow window of opportunity in the EU and potential development timelines to determine whether consumer demand will generate enough profit to warrant investment in EREV powertrains. Notably, EREV powertrains may be a more future-proof option for OEMs than PHEV powertrains because they combine a BEV platform with a small ICE-powered generator that is not connected to the drivetrain. Unlike the European Union, the United States has no zero-emissions requirement in place for new-car sales. At the federal level, Environmental Protection Agency standards tie compliance bonuses to electric driving ranges, which indicates a potential advantage for EREVs over PHEVs.12 For example, an EREV with a range of at least 70 miles could receive a 65 percent bonus in compliance standards, while a PHEV with a 25-mile range could receive a 30 percent bonus.13 The California Air Resources Board’s Advanced Clean Cars II rule does mandate 100 percent electrification in new cars sold by 2035, but one-fifth of those vehicles could be PHEVs or EREVs, and manufacturers could receive full credit for each vehicle with an electric range of at least 70 miles [https://www.mckinsey.com/industries/automotive-and-assembly/our-insights/could-extended-range-evs-nudge-more-car-buyers-toward-full-electric](https://www.mckinsey.com/industries/automotive-and-assembly/our-insights/could-extended-range-evs-nudge-more-car-buyers-toward-full-electric?ref=nicholasjohnson.blog) ![](https://nicholasjohnson.blog/content/images/2026/02/data-src-image-2abc97f1-0f30-44ff-8324-2bbd227ca3b8.png) **Options for Generators that are small enough to keep an EV charged while in motion, even when parked.** [https://fusionflight.com/arc/](https://fusionflight.com/arc/?ref=nicholasjohnson.blog) Should EVs with small petrol range extenders be more popular? I recently read about and watched some videos on the new Mazda MX-30 R-EV in Europe. The new version fixes the range issue that the previous MX-30 had (I still don't understand why they designed the MX-30 with only 160 km of range) by using a small rotary engine that charges the battery while driving. Now, this specific model, the rotary engine, isn't that great in MPG, but there has been another car that has used a similar range extender with a non-rotary engine - BMW i3. Why isn't this concept more popular? This type of system could be a great way to be a middle ground in the EV/Petrol market, offering benefits that offset both sides. You can build an EV, reducing the duplication/complexity of full-hybrid systems that use both gas and electric engines in tandem. As a full EV, you need little maintenance and less to go wrong. Range extenders can be small, lightweight engines that only charge the battery and boost range. You get the electric benefits of an EV at short distances, but also the range advantages of gas power over longer trips without needing to recharge. This seems to address some of the range anxiety that arises when using charging networks. By using a range extender, you can opt for a smaller battery, saving weight and cost (somewhat offset by the need to build a small engine for the car). Now, the MX30 has other issues: crazy suicide doors, a small back seat, and the R-EV isn't even available in North America - but I would die for just a regular car with a similar concept. [https://www.reddit.com/r/cars/comments/1e1sz89/should\_evs\_with\_small\_petrol\_range\_extenders\_be/](https://www.reddit.com/r/cars/comments/1e1sz89/should%5Fevs%5Fwith%5Fsmall%5Fpetrol%5Frange%5Fextenders%5Fbe/?ref=nicholasjohnson.blog) **Conclusion** As of now, no generators on the market meet this exact need. We don't need a gas-powered engine, but a generator that can produce about 30-40 kW of continuous power from a small fuel supply. But adding the complexity of the system makes no logical sense compared to a pure battery-electric vehicle for most of the US market, and especially in Europe and China, where distances traveled are shorter. That said, niche markets like trucking and large SUVs could benefit from range extenders.The overall market consensus, though, will be sentiment. How do you sell this as a better, cooler vehicle with more capabilities than the current status quo of pure battery-electric vehicles? Behavior is not based on numbers or logic. ### A Founder’s Reflection of 2025 URL: https://nicholasjohnson.blog/a-founders-reflection-of-2025/ Last updated: 2026-01-01T20:03:18.000Z I started Orange in 2020, going straight from CTO of LYT.ai into founding the company and diving headfirst into building our first product, Orange Outlet. We were backed early by an extraordinary group of mission-driven investors and advisors: Marc Tarpenning and Martin Eberhard (Tesla, pre-Elon), Sven Thesen (Nobel Prize–winning climate researcher), and Marc Geller (Plug In America board member). They weren’t just capital—they were conviction. They gave me the push to go all in. Over five years, Orange grew into a national EV charging company. Along the way came the inevitable growing pains: new hires, new ideas, new funding rounds, and eventually institutional capital. The scrappy, garage-style startup I started had to evolve into something more corporate. That transition fractured the team in ways I didn’t fully anticipate. By 2025, most of the original group I started with in 2020 had moved on. In 2025, we closed a large Seed round led by MRV and Climactic and brought in more experienced leadership. That experience came with friction. My “get shit done” instincts and scrappy execution style clashed with more process-heavy approaches. After months of team conflict, I decided to step away from Orange, taking with me five years of hard lessons and the freedom to start fresh. Not in a market I fell into, but one I would choose. That freedom mattered. Without the constant pull of operational fires and people issues, I spent four months in NYC in 2025 exploring ideas, industries, and relationships more deeply than I had in years. It gave me space to observe, learn, and reconnect with why I built in the first place. At the same time, the EV charging industry took a hard hit. The Trump administration rolled back incentives, while California’s increasingly complex regulatory environment raised the bar for selling and deploying chargers. As tax credits wound down, EV sales spiked briefly in Q3 before cooling. OEMs like Ford cut or paused EV production lines to regroup. Meanwhile, Europe and Asia—especially China—continued pushing forward on energy independence. Zooming out, the energy landscape itself is shifting fast. Data centers are driving electricity demand into a new stratosphere. Incentives for renewables are shrinking, even as compute demand explodes. Something fundamental is re-forming at the intersection of energy, infrastructure, and software. At the same time, the way software is built has changed completely. For decades, software was rigid - every line scrutinized, every edge case anticipated, complex logic hard-coded into brittle systems. In 2025, that model cracked. AI and machine learning began handling complexity directly, trained on real-world data using large foundational models—many of which are open-sourced and rapidly scaled. Software is becoming adaptive, contextual, and far closer to reality than we ever imagined. That shift sent me on a research journey across energy, healthcare, data centers, freight forwarding, trading, and work communication—searching for places where AI can create real leverage for humanity. The hype has been intense. Valuations have been wild. Some are calling this the biggest tech bubble since the dot-com era—but there’s a key difference. In the 1990s, networks were built before users had the devices or bandwidth to use them fully. Adoption lagged behind infrastructure by years. Today, that gap no longer exists. People already have the devices. The networks already move data at scale. And we still can’t build data centers fast enough to keep up with AI’s computational demand. The acceleration is real—and it’s only beginning. That’s why, in 2026, I’m doubling down on finding people and industries that benefit from AI beyond a simple chat interface—through streamlined data pipelines, continuous learning systems, and deeply integrated workflows. The next 12 months will lay the groundwork for one of the most significant technological shifts humanity has seen. And whether we admit it or not, the world knows it. As with many of the projects I’ve gravitated toward - clean energy, autonomous systems, infrastructure -I’m excited to spend the next year at the frontier of what’s possible when we rethink how technology integrates into humanity’s most complex systems. 2026 feels like the beginning of something big. If you want to keep up with the progress of this exploration sign up to follow along at ZestyLabs.org ### Zero to One Through Great Customer Discovery URL: https://nicholasjohnson.blog/0-1-is-just-great-customer-discovery/ Last updated: 2026-01-03T21:45:52.000Z Over the last several months, I've been helping several other founders who are early or have had to pivot from their original idea. Each time I encounter the same friction or resistance from founders, I have been there myself, facing the pushback of talking with customers and asking for money. > Customer discovery is probably the most essential part of building a business. Still, it is often misunderstood, and the processes around it are often not strong enough to hold each other accountable. Too many founders speak to too few people and assume that their feedback is sufficient to build a business around. The problem is, this is neither true nor scalable. You need hundreds, if not thousands, of customers who have similar pain points that you can aggregate into a product. This means that customer discovery needs to be a process that collects data from multiple sources. Your job as a founder is to listen, understand, and aggregate that into a product that is compelling and creates value for the end customer, where the pain point is strong enough for them to pay a significant amount of money to have that problem solved, hopefully by your company! Many startup founders I've spoken to over the past several months lack a straightforward, methodical process for aggregating data and unifying it into something useful. Often, listening to one or two customers or design partners to build a product. This results in a glorified free-consulting job for the design partner. Mainly because many founders I've spoken to are also hesitant to discuss money early on in conversations, fearing they might scare off potential customers. Many startup founders I've spoken to over the past several months lack a straightforward, methodical process for aggregating data and unifying. The founder's discovery is relatively simple, but it is a painful process of emailing, calling, and outreach that most technical founders dislike, simply because that's probably why they became technical. **You need a precise balance between:** 1. Technical understanding of what's capable 2. What the customer is saying 3. What you believe you're going to build 4. The ability to execute on it Meaning you still need someone to build the thing once you figure out what it is, but often, figuring out what to build is hard. How many times have you heard "I want to start a company, I just have never had teh right ideas"? One, please don't start a company if you think the idea is what matters; it's the desire to solve real problems others haven't yet solved. It's going to be hard because if it were easy, someone would have done it already. I want to spend some time demystifying the customer discovery process, also known as finding product-market fit. They're technically different things. But without one, you can't find the other. A clear understanding of Product-Market Fit occurs when customers continually buy more of what you sell or keep asking for it. This is the best example of Product-Market Fit. You'll know it when you see it, because customers are clearly embracing what you've built. Because they love using it, it solves a real pain point and adds significant value to their workflows, lives, or businesses, or it's just really addictive. - Please don't become a drug dealer like TikTok or Facebook. One of the key problems, though, is that customer discovery doesn't look the same for every company. Nor is there an easy playbook you can simply copy and paste for each business, because customer discovery is deeply ingrained in the industry, the customer, and the methods for engaging them. Customer discovery looks very different for B2B and B2C companies, and even more so depending on if you're a research project (such as a nuclear reactor company, a fission company, or an AI company doing customer products), or if you're doing an AI company that's doing deep learning and models. Your customers are vastly different across these businesses, so customer discovery will look significantly different because the addressable markets themselves are distinct. I'm not sure why so many MBAs believe they understand this, but when I observe them in action, they cannot develop a process that transforms the data they collect into a refined product. They are aware of this, but when I observe them executing, they have no means to develop a process that converts the data they collect into actionable insights to inform their decisions. This is why technical capabilities and strong intuition are critical for turning that data into action. --- Clear customer discovery should help answer these questions with excruciating levels of personal and intimate detail about the industry, the buyer, the problem, the user, and the overall market that you're addressing. 1. Understanding your market. 2. Understanding the problem. 3. Understanding the economics. 4. Validating your data for a product. 5. Defining a clear GTM and repeatability. 6. Understanding how your customers buy. Step one: Identify a market opportunity. Define the pain points and talk to customers. Determine whether your customer is a buyer, a decision-maker, or involved in the buying process. For B2C, it is as simple as running an ad. Some of the biggest brands you've ever heard of literally didn't even have a product when they first started advertising They had an idea of what they wanted to sell, but didn't have any supply chain/product figured out. All they had was an idea, and they marketed it based on strong intuition. B2C companies can do this because the feedback they receive from marketing tells them whether it's a good investment to actually start building the product. Liquid Death is the stupidest company you could imagine. "It is canned water." Going to an investor and telling them you want to build a $100M business at the time of writing this, but all you're going to do is take water and put it in a fancy aluminum can. Oh, and your branding is dark and scary. Seriously, I want you to really understand how dumb this idea sounds when you say it out loud. But the founder had understood the problem. He'd been to Warped Tour and seen monster cans labeled 'water' for the band members. Because no band member is downing Monster while they're singing their top songs in front of 10,000 people. There's a market of people who feel stigmatized for drinking water from plastic bottles, and who believe the overall environmental impact is significant enough that, if you could band behind it, people would buy your product in support of a message that is greater than the sum of its parts. How did he do customer discovery? He spent a few thousand dollars on a video [linked here](https://youtu.be/EeRADNpdKD4?si=oQ7yaY6OTrBS34ot&ref=nicholasjohnson.blog) that advertised Liquid Death in a way that made it stand out from the boring water companies he was competing with at the time, he had no product, just a cheap rendering that somebody had done for him off Fiverr of artwork he had created for the Cans, an actress who was in Hollywood looking for work, and a simple low-budget advertisement that he put online that racked up millions of views. That was enough validation to invest in the first shipment, an order of Liquid Death. Canned still and sparkling water from the Austrian Alps by partnering with energy drink companies' existing supply chain. He was able to go to market cost-effectively and efficiently, and use the momentum from the Simple campaign to sell out. Liquid Death's most valuable asset continues to be their ability to market to an audience that loves the differentiating factors of canned water with edginess that makes you feel like you're a part of a movement, not simply buying water off the shelf and seeing the trash pile up in our oceans. Seriously, they sell a tall-can of water for $3.25 USD. Yet so many founders I talk to want to do this the other way around. They have this product they want to build, and they believe people will like it. Everything from canned peptides to new energy drinks. I can list the number of friends who have spent thousands, if not tens of thousands of dollars, on a product that just fills their garage because they never conducted proper customer discovery or understood what it took to bring it to market. Mentally, a business comprises several key components, and your role in customer discovery is to understand each part of the business and how it will operate to achieve success. The best companies know that all of them need to overlap cleanly to reach escape velocity. Imposter syndrome or fear of failure. This is probably the biggest source of pushback I see from founders, and sometimes from myself. I call it the failure to launch or the failure to make the phone call. I can tell you it's taken me years to overcome this initial inertia. At my first company, we had hit a dead end. We had talked to all the OEMs and realized: 1. Autonomous cars were decades away 2. OEMs were extremely difficult to sell to 3. They weren't interested in a 3-person start-up out of a Sunnyvale building connected vehicle technologies when they themselves couldn't figure out how to build infotainment systems to compete with Tesla, and most management was too old to understand autonomous cars. We hit a dead end; connecting cars wasn't that valuable. Over the next several months, I made numerous phone calls to various industries to determine what would make the most sense. We had a whiteboard in our Sunnyvale home living room that filled the entire wall, and on it we had ideas we thought sounded fun. Personally, I'm still sad that we never built the trash-cleaning [Wall-E](https://youtu.be/7oVSaUWeKt0?si=jMOMMROI8DhieBCM&ref=nicholasjohnson.blog) that would have gone around the sides of the road and picked up trash. Something just sounds very fun about it. But it was clear that no market or go-to-market motion would make that product successful in 2017\. However, several phone calls made it clear that there is an interest in improving traffic management and innovation related to autonomous vehicles within cities. Therefore, after several calls to the cities of San Jose, Chicago, L.A., Miami, and Fort Lauderdale, Florida, we found that innovation teams in those cities were exploring products to optimize the flow of all types of vehicles. Some of these were: 1. At the time, we learned that scooters were showing up everywhere, and a need to track them so they could better regulate scooter deployments throughout the city, to not end up with high messes of bikes and rideshare scooters everywhere 2. The advent and acceleration of ride-sharing is pushing cabs and making it hard for public transit to compete. 3. The fear was because at the time, the number of autonomous vehicle companies, including Tesla, was promising that AVs were a year away, that transit would change so drastically that it would become obsolete. These fears helped us understand what our customers were making decisions about. Next, we had to know how our customers made decisions. I won't lie, this took us a lot longer and was a slow, painful process. If you had told me I was going to sell the cities and how hard it would be to sell because of the number of people involved in the buying process, I would have turned around and never done it. We don't do hard things because they're hard. We do hard things because we thought they would be easy. Sometimes the best advantage someone has to do something challenging is the ignorance of how difficult it really will be. "Stay hungry, stay foolish" - Whole Earth Catalog. I want to double down on this last point because, as you get older, it becomes more difficult. But as a 20-something out of college, your optimism often overpowers the pain points that will come ahead. The best founders maintain a level of childish optimism throughout their lives regarding ideas and challenges, without jading them into pushing off things that could be done because they have too much information. Because of this customer discovery and the number of phone calls I made, many of which didn't lead to anywhere, we were able to aggregate around how we would go to market and get opportunities within those cities who were able to help us get to the next level, which is access to traffic lights, data, and the things we needed to build the product that became Lyt. This is just one example that I personally have. I've done this several times now. Again, an Orange, realizing that sometimes the product we had wasn't correct, or pivoting and understanding what it took to actually get to market. Fundamentally, we still made bad decisions along the way because, as we grew, we had a rigid team that was not willing to learn from the feedback we were receiving. Side note, people from corporate America generally suck at testing new ideas, validating them, and moving on. Most people believe their job is about executing an idea from start to finish. Because that's what they've been trained to do in their organizations. This is a massive problem I see: startup founders who have already had jobs at large companies with cultures that aren't like a start-up's. > Also, why were all the Magnificent Seven started by young 20-something-year-olds who have continued to learn and advance the startup mindset that allows them to execute quickly, explore the world, kick out ideas, and try new things? The number of mistakes they've made is greater than the number of successes. However, people are often afraid of making mistakes and/or learning that their idea is wrong due to ego or self-reliance, which causes them to ignore early signs and delay pivoting. By the time they hit a wall, it's too late. One of the best examples of a B2B play that I've ever heard of in history, which created one of the wealthiest individuals on the planet, [is Bill Gates' pitch to IBM](https://youtu.be/9nfgRf2A0Tc?si=6k-JKXJWAHHAk4uQ&ref=nicholasjohnson.blog). At the time, they were consulting on software solutions for other companies within the ecosystem of computer technologies or IT systems. They were dropouts, but they wanted to be part of the revolution that was unfolding during the 1970s and early 1980s. They had no idea what their actual product would be, and the industry was undergoing rapid change. IBM, which had been the lead player in information technology and computing for decades, had the opportunity to learn, pivot, and execute better than anyone else. However, as in actual history, the people at that stage were not rewarded in those companies for taking such risks, and therefore made a fatal error. > The reason is that these people have continued to learn that failure is not a big deal. In fact, in some cases, it is the best way to learn. One of my favorite movie quotes is, "A man often learns more from his failures than his successes. The trick in life is to not make a habit of them." Bill Gates pulled off a remarkable move by simply partnering with IBM. This company had all the money, all the talent, and everything it needed to essentially build what Microsoft would become. He negotiated a deal with executives who lacked insight and were not rewarded for taking risks, which ultimately led to the handoff of a licensing agreement that made Microsoft the most valuable company in the world. This agreement, however, was overshadowed by their subsequent struggle to compete in a commoditized hardware market over the next 20 years, which ultimately led IBM to exit the market. We want to emphasize that IBM remains an exceptional company and has, over time, developed a core understanding of B2B and networking to create products used by many corporations today. However, they themselves missed out on becoming the next Apple or Microsoft simply by giving a 20-year-old access to a perpetual license to software developed for IBM hardware and giving that software a go-to-market (GTM) through IBM's brand and scale. However, it is also worth noting that Bill Gates and the team at Microsoft did not recall a software product that became standard and ran IBM's computers for the next several years, as well as many other computers that entered the market at the time of this deal. Bill Gates's ability to understand the changing ecosystem and the opportunity at hand allowed him to negotiate an agreement before actually writing any code. In fact, his actual value was knowing the guy who had written the code, who was a passionate engineer but had no access to this business model, nor the capital needed to scale it. The story goes that Microsoft bought the original BASIC software, which became a key product offering sold to IBM for the next several years, for $50,000\. And with every subsequent computer sold, Microsoft charged a licensing fee for the software, making millions upon millions of dollars and allowing them to develop what became Microsoft's OS, which we still use to this day. If you don't see the trend here, it's that some of the largest and most impactful companies have made their mark by trying to sell, negotiate, and learn about the market deeply before making an actual investment in the product. This meant they had the resources to continue on another day while they figured out how to get the most leverage from the opportunities in front of them. --- I hope this is a helpful example to get your brain thinking about how you, as a founder or an executive, entrepreneur, or entrepreneur inside an organization, properly leverage resources to understand what you need to build based on customer discovery before you actually invest in creating it. **Exceptions to this concept.** Much in life is not black or white. This is an extremely valuable framework to consider when undertaking your next endeavor, and the most likely way to achieve customer discovery and product-market fit is to go out and talk to as many people as possible to identify opportunities you can leverage to build a business around. There are exceptions to this rule, and many of the most prolific and great products we use every single day were not developed using this strategy. I'm going to talk about two products that we know well in the Western world: Gmail and Slack. Two software tools that are used by many people every single day to communicate, and how they both became products, not because of an overall initiative from above, or a genius idea, or customer discovery, but because they were used by the end-user who was creating them to solve a problem that they faced every single day. It is the exception to the rule. Suppose you are the ideal customer, and there are many people like you worldwide who have the same problem. You can build the proper tool over time that excels at solving that problem, and you have access to a way to scale it and get it to market. In that case, you can succeed without completing the first part of what we just discussed, which involves intensive outreach and customer discovery. Instead, you focus on the things you care about in the product, assuming that you are the norm. Now, there's an intentional risk associated with this, but it has worked out for many companies in the past, as exemplified by Gmail and Slack. Let's start with Gmail. It's been around longer and is the fundamental way most people think of email, but it didn't start out that way. As Google scaled and grew, it hired many great software developers. [You can learn more about Gmail's history, including how it was introduced into the Google ecosystem and later brought to market. ](https://en.wikipedia.org/wiki/Gmail?ref=nicholasjohnson.blog) The Cliffs-Notes are simple: An engineer at Google was working on a product for himself to better manage the daily emails that came through between teams. He wanted a way to better message people, so he built Gmail for himself. He then let others start signing up on Google. People flurried all over the company to get access to this new tool. It caught the attention of the Google executives, who then quickly spun it out into a product as part of the Google Suites. This is a perfect example of not conducting customer discovery, but rather solving a problem that you face every single day, which many other people also experience. That your passion and execution are good enough to become something other people will cling to and use. Many products often use this approach to grow silently within an organization. Facebook utilized academic institutions, while Gmail leveraged Google's internal needs to validate, define, and develop the product. By the time the public gained access to it, it was almost a fully featured and polished product. The feedback they received internally was a good generalization of the views of all other users outside the organization. Next is Slack. Slack's story is even more interesting. During the early dot-com bubble and the early 2000s, a group of small founding developers was working on a video game world that ultimately flopped. During the launch attempt for this product, they built Slack internally to help them develop it. They themselves needed a better solution than in-line sharing in emails to manage the product's and game's assets, so they could communicate effectively and build what they thought would become a massive success. The product they built internally to help develop the game that failed ultimately became their great success. These are two exceptions to the above rule of customer discovery. While, like I said, the world's not black or white. These are exceptions and are rare in the reality of building businesses. In fact, most companies built this way (where the product is overhyped, spends too much time in development without talking to real customers at scale, and has a process for incorporating feedback into the product's decision-making) fail miserably. There are more examples of companies that attempt to do what we just discussed that fail than succeed. Just this last year, a few AI pins failed spectacularly by staying in stealth mode for too long, without customer feedback outside their bubble. > In fact, most companies built this way (where the product is overhyped, spends too much time in development without talking to real customers at scale, and has a process for incorporating feedback into the product's decision-making) fail miserably. I want to point this out, as many people will cling to it, explaining that their idea is so good or that they understand it so well. I've yet to see this, but I know better how the market works out. --- Lastly, bad advisors plague the start-up ecosystem like never before. People who have massive egos and think they know best because they have either been successful or have an inflated idea of themselves. They will often give young or new founders advice on their ideas, saying they understand them better than the person building or the person with the idea. That said, there are bad ideas, and that is just life, and many people shouldn't start companies; it's harder than people make it out to be. At the same time, no one knows what will be a wild success. So teh advice on an idea is not practical, especially if you only have one data point. What great divers do is help people focus and run a great process to complete customer discovery so they can answer teh hard questions and build a strong intuition about what the company should do to get to the next step in building from zero to one. ### Three Quarters of Growth: What Orange Learned Building EV Charging for Multifamily URL: https://nicholasjohnson.blog/three-quarters-of-growth-what-orange-learned-building-ev-charging-for-multifamily/ Last updated: 2025-08-16T06:59:36.000Z **Context:** A deep dive into Orange’s last three quarters—covering GTM execution, revenue growth, hardware challenges, and the unique dynamics of selling into multifamily properties. Focuses on business model design, utilization, connectivity, and why Orange’s approach is fundamentally different from traditional EVSE providers. **Post Body:** When we started Orange, we weren’t trying to build just another EV charger. We set out to re-architect how EV charging works for the places most Americans actually live: multifamily housing. In the past nine months, we’ve grown bookings nearly 400% year-over-year, expanded revenue over 200%, and more than doubled our deployed footprint—while proving our business model is not only scalable but necessary for electrification. Here’s how we did it—and what we’ve learned. --- #### 1\. **The Multifamily Problem is a Systems Problem** Over 52 million Americans live in multifamily housing. By 2030, 27 million EVs will be on the road. But today’s infrastructure wasn’t designed for this. Properties don’t have the energy capacity, budget, or staffing to support large-scale EV deployments. Most solutions assume: - There’s unlimited electrical capacity - Drivers have assigned parking - Buildings can absorb $5,000–$8,000 per stall in install cost - WiFi/cellular always works That’s fiction. Orange built a system grounded in the reality of multifamily: - Plug-in chargers that require no panel upgrades - Software that optimizes energy usage building-wide - A mesh network that operates without WiFi or cell - Hardware and install for under $2,000 per stall --- #### 2\. **Our GTM Motion is Built for Volume, Not Hype** We’re not doing $100K pilots. We’re doing 50-unit deployments. And we’re doing it repeatedly. - Booked 11,041 chargers in 2024, up from 567 the year prior - Achieved 70%+ quarter-over-quarter growth in revenue and backlog - Installed base grew from 220 units in Q3 2023 to 1,187 in Q3 2024 - Active drivers rose from 52 to 248 in the same period - Energy delivered jumped 700% This isn’t just growth. It’s repeatability. We closed deals with the top 10 multifamily operators in the country, deployed in 120+ properties, and saw bookings from expansion increase 269% YoY. --- #### 3\. **Why Our Business Model Works** Most hardware companies get crushed by: - Capex cycles - Service contract overhead - Installation complexity We sidestep those: - Hardware is simple and modular - Property owners don’t pay upfront for software - Drivers pay per kWh ($0.08), which covers network and platform - No cellular or backend IT systems required on site - Utilities get real-time usage data through our submetering platform Result: hardware gross margin scales, energy revenue compounds, and we control the user experience end to end. Our ARR engine: - \~$800 per charger (hardware revenue) - $240 per active EV driver annually (energy usage) - $100M ARR target with just 1.4% of multifamily EV drivers --- #### 4\. **What’s Hard—and What We’ve Solved** Multifamily is brutally difficult: - No two buildings are alike - Retrofits are messy - Property managers don’t want one more thing to deal with We solved it by: - Pre-scoping buildings with proprietary tools - Standardizing hardware to 12/8 AWG configurations - Offering zero-maintenance devices - Allowing billing to be added to tenant utility bills via our PCE submetering grant And, most critically: we proved installation can scale without breaking the property or the budget. --- #### 5\. **How We’re Positioned to Win** The EV charging industry is full of overbuilt, overpriced, brittle solutions. Companies that raised $60M+ are stuck with 5 installs and bad product-market fit. We grew past them with: - A $1,750 installed cost vs. $7,000 industry average - Fully offline-capable charging sessions - No networking gear required—even in underground garages - Seamless mobile app that works in airplane mode CalGreen 2024 requires 50% of parking to be EV ready. Orange is the only company that can meet this with one unified system. --- #### 6\. **The Next 12 Months** We’re on track to reach cash flow positive by Q1 2026. What’s coming: - Production ramp of OrangeOS with edge-based energy management - Launch of V2G-capable hardware with Range Energy (pending $2.7M CEC grant) - Expansion of utility billing integrations - Self-service portal for property deployment We’re building not just a charger—but the foundation of a distributed energy utility. --- #### Final Thoughts In the last three quarters, we built the most scalable EV charging company in multifamily. Not because we raised the most, but because we built the smartest. We optimized around the market reality, not the fantasy. We’ve grown 3x+ YoY, proven a profitable business model, and laid the groundwork to own one of the most important energy transitions of the next decade. And we’re just getting started. --- For partnerships, data requests, or press inquiries: Nicholas@orangecharger.com ### Too Many Cooks in the Kitchen: Baking EV Charging URL: https://nicholasjohnson.blog/too-many-cooks-in-the-kitchen-baking-ev-charging/ Last updated: 2026-01-01T20:15:00.000Z ![](https://nicholasjohnson.blog/content/images/2025/09/EVSE-Code-Regs.png) ## **⚖️ Navigating the Complex Standards Landscape of EVSE** The biggest challenge facing EVSE innovation today is not a lack of technology, but rather the overwhelming number of agencies simultaneously attempting to regulate it. The adjacent Venn diagram illustrates this crowded space. Each organization serves a purpose, but when its standards conflict, overlap, or overreach, it introduces cost, delay, and sometimes even drives companies out of the U.S. market entirely. ### **NFPA 70 - National Electrical Code** While published by the National Electric Code (NEC), NFPA 70 is a "model code" that is then adopted by states, typically with minor changes each year. These changes take years or decades to trickle through, meaning that every state is essentially an Island with regard to codes. This means that EVSE standards imposed here create a wide range of adoption variations among states, further creating a burden on installation and operation for national distributors and companies looking to standardize and reduce costs. These increasingly complex changes add cost and confusion. Worse still, the NEC is positioning itself as the definitive authority on EV charging, which often shifts from safety concerns to political infrastructure relevance and away from product safety. Please start treating EVSE installations like the standard circuit they are. ### **Other NFPA Standards - National Fire Protection Association & Local Fire Codes** NFPA’s core mission is fire safety. That is valid and vital; however, their expanding influence over EVSE installation requirements includes energy power shutoffs and added signage. EVSEs are fundamentally electrical appliances, and their fire risk is not unique relative to other household or commercial electrical loads. NFPA should limit its oversight to fire-related design standards and defer to existing frameworks for electrical and automotive safety elsewhere. Prescribing power levels, communication protocols, or system architecture without deep domain knowledge risks confusion, cost inflation, and worse—stagnation. An example that stalled all charging installation in San Francisco last year was the city fire department requiring fire sprinkler upgrades for the installation of EVSE in parking garages from a 1.5gpm to 2.5 pgm, often requiring a new water main, and hundreds of thousands of dollars of new fire sprinklers to install $80-100k in charging. ### **SAE - Society of Automotive Engineers** SAE plays a critical and appropriate role in the EVSE space—defining plug formats, communication protocols, and vehicle-side interface standards. This is necessary. The EVSE’s primary function is to communicate safely and effectively with the vehicle. SAE’s leadership on ISO 15118 and related V2X protocols reflects its understanding of what vehicles need to function safely and reliably. This standardization helps ensure global compatibility while supporting long-term innovation and growth. SAE has a long-standing track record and established process for developing standards that promote interoperability among automotive companies, while also setting minimum safety requirements for vehicles when adopted by national homologation organizations. ### **UL - Underwriters Laboratories** UL is the anchor for safety at the device level. Its job is not to prescribe how EVSE should be designed, but to ensure that whatever is brought to market meets minimum safety and operational reliability thresholds—across leakage current, arcing, temperature, and more. UL’s standards strike an essential balance: enabling innovation while protecting users. UL 2231, UL 2594, and efforts like UL 1741 for V2X are examples of standards that evolve with the market, rather than trying to control them. While UL is not perfect, the process for defining and maintaining product-level safety standards is where EVSE safety should be determined. ### **Building Codes** Building codes, particularly those outlined in Title 24 and CALGreen in California, are increasingly mandating the readiness of electric vehicles (EVs). This can be beneficial, but it can also be hazardous if it is misaligned with the broader ecosystem. Building codes should not dictate how safety requirements are met or what qualifies as an EVSE or access to energy to charge cars; instead, they should rely on standards established by organizations such as UL or SAE for those definitions. These decisions need to remain flexible, focused on outcomes (readiness and access), and not prescriptive to specific power levels or technologies. They should treat EV charging as a black box and ensure the safety of the infrastructure downstream of any circuit, as well as precise safety requirements for construction. Reach codes also touch EVSEs and often specify electrical details. For example, 40-amp circuit requirements add to excessive power budgets and added copper costs that will never be used. ![](https://lh7-rt.googleusercontent.com/docsz/AD_4nXfM87ewGpjrK2Ykaw-O0x91rhytji_tAFYtpExjZsyELyTlyOKThG_NGgVauMa6aPMvxFWvLsbppD0lwCHYc3mtFc04wfrKehHOMwEyPJx2rJNYFFVp3eAmAtBhAw2kM0EScXbW?key=yhfKR4VUnbCxZFcIhl1Zwg) ### **The Federal Access Board (ADA)** As of this writing, guidelines have been established, but not national standards, for charging accessibility. Instead, it’s a state-by-state thing. However, docket ATBCB-2024-0001 will turn guidelines into rules. Some of these rules present barriers not to persons with disabilities, but to accessing charging facilities at all, making it impossible to retrofit charging into existing parking lots legally. ### **NEMA - National Electrical Manufacturers Association** NEMA represents manufacturers of electrical equipment and plays a significant role in shaping technical standards and code proposals, often influencing the language and interpretation of the National Electrical Code (NEC). In the EVSE space, NEMA’s members include companies with vested interests in traditional infrastructure and hardware-heavy solutions. While NEMA provides valuable insights into component safety and product interoperability, its advocacy sometimes reflects business models that favor large-scale, high-cost hardware deployments, which may be at odds with emerging lightweight or distributed charging solutions. When NEMA pushes for prescriptive hardware requirements or discourages flexible architectures, it risks entrenching legacy interests over market innovation. NEMA’s input must continue to focus on safety and interoperability, rather than using standards as a lever to preserve incumbent market positions. ### **EPIR - Incentive Programs** EPRI plays a supporting role in shaping how utilities and energy systems adapt to mass electrification, but has increasingly influenced funding eligibility through technical specifications. While their research into grid impacts is valuable, their involvement in incentive program criteria has led to prescriptive technology mandates, such as requiring OCPP 1.6, OCPP 2.0, or ISO 15118 for EVSE eligibility. These protocols have no bearing on safety and instead prescribe specific architectures for how chargers must communicate, often favoring one design approach over others without regard to cost, user needs, or deployment complexity. EPRI’s role should remain focused on grid readiness and infrastructure modeling, not narrowing the field of innovation through backdoor technology mandates embedded in public funding criteria. ### **💸 Incentive Programs and Political Prescription** A final, yet increasingly powerful, influence on the EVSE ecosystem comes from incentive programs—often created as political infrastructure by state and federal agencies with good intentions but poor technical alignment. Many of these programs now require features such as OCPP 1.6/2.0 or ISO 15118, minimum delivery (kW) requirements that require all ports to be able to deliver 100% all the time, ensuring the electrical system is vastly oversized to the actual needs for eligibility, regardless of project scale, customer use case, or economic viability. This prescriptive approach turns public investment into a rigid solution matrix, rather than leveraging the private sector’s creativity and technical depth. At their best, these programs should support outcomes—**access to safe, cost-effective charging where vehicles are parked**, not dictate the protocol or plug that delivers it. ### **🚧 The Real-World Implications** Many EV owners never install a wall-mounted Level 2 charger. They plug into a standard outlet or a NEMA 14-30 dryer plug with a mobile connector and charge their vehicle overnight just fine. When we start layering prescriptive GFCI requirements, power minimums, and V2X constraints into code or incentives, we risk outlawing cost-effective, user-friendly charging solutions that already work. The recent changes to 625.4 will push more homeowners to install outlets that they pretend are not for EVSE. Yet, hardwired EVSEs are easier to install correctly. As the CEO of Orange, I’ve seen firsthand how conflicting requirements from these organizations slow innovation. Even when we meet safety standards, incentive programs, or fire code interpretations may add yet another layer of conflicting logic that disqualifies products like the Orange Outlet in favor of high-power wallboxes. That’s not progress—it’s calcification and stagnation of innovation. ### **🧩 Conclusion: A Call for Collaborative Restraint** Every agency in this diagram has a role to play—but no one agency should overstep its expertise. NEC, NFPA, UL, SAE, and local building code bodies must respect one another’s domains and align around a shared outcome: enabling scalable, safe, and affordable electric vehicle (EV) charging infrastructure. The EVSE of the future will support not only charging but also grid integration, energy optimization, and resilient power. If we over-regulate today, we strangle that future before it arrives. Let’s allow innovation while ensuring logical, clear safety standards from organizations like UL and SAE, who have strong track records of evolving with industries vs stifling them. ### Why 🔥 Burnt Outlets Are Dangerous URL: https://nicholasjohnson.blog/why-burnt-outlets-are-dangerous/ Last updated: 2026-01-01T20:14:32.000Z When installing EV charging, several critical factors must be considered to ensure the infrastructure's lifespan. The first is how to provide power to a parking space, which can be achieved by using an outlet or a hardwired charger. ![](https://nicholasjohnson.blog/content/images/2025/07/IMG_8338.jpg) Outlets are the most affordable and easy way to switch between charging options for drivers who might have a J1772 or NACS cable shipped with their vehicle. They’re also the most common way to charge an electric car, with over 75% of EV owners plugging into a 240-volt outlet at home a few nights per week, effectively waking up to a charged vehicle similar to your phone today! There are many types of NEMA outlets worldwide, each providing different benefits for various applications. Many people install an overkill NEMA 14-50 or use one of the many adapters to plug into what they already have wired to their homes. #### All the NEMA types a Tesla mobile connector can plug into. ![](https://nicholasjohnson.blog/content/images/2025/07/image.png) The issue is that not all Outlets are created equal, and most 240-volt outlets customers buy are only designed to be plugged into a dozen or so times for appliances like washers, dryers, stoves, refrigerators, etc. With lower duty cycles, these outlets wear out and pose a fire risk as they get used daily for EV charging. In Orange’s first year in business, two employees reported that the NEMA 14-50 outlets wore out and melted. ![](https://nicholasjohnson.blog/content/images/2025/07/ev-charger-blew-out-woke-up-to-smoke-and-found-melted-v0-9tr8u68iajx91.jpeg) ![](https://nicholasjohnson.blog/content/images/2025/07/IMG_0094.jpg) ![](https://nicholasjohnson.blog/content/images/2025/07/the-14-50-car-charger-clippercreek-melted-is-it-the-outlet-v0-wotymb9hf9yc1.jpeg) ![](https://nicholasjohnson.blog/content/images/2025/07/Metleted_NEMA_1450.jpg) The outlet melted due to wear and tear, as it was a low-cost Home Depot $13 consumer-grade device. The same issues occurred, but here it also melted the cable to the charger. Their extension cord burned out, as it lacked temperature sensors to inform the onboard charger to stop pulling current. Due to the high resistance, the contacts get extremely hot from extended current draw, leading to complete melting. This happens as the contacts inside, often made of cheap brass for consumer-grade outlets, wear out and apply less force to the plug after repeated use, allowing resistance to build up between the contact and plug. This resistance generates heat and often poses a fire risk. To reduce the risk, NO ONE should install anything less than an industrial-grade outlet for EV charging applications; these outlets are more expensive and designed for more duty cycles and high continuous current loads for industrial applications. They start at about $98 per plug for most applications and increase to $200 for high-end hospital applications, where reliability is critical for life. Then there are Orange Outlets. Orange Outlet exceeds the highest standards by being the first fully engineered Outlet, from day one, designed for use with Mobile connectors and exceeding UL 485 receptacle standards. Each outlet contact is made of copper beryllium, a material that exhibits a higher spring force and is more rigid, thereby preventing deformation over repeated use. Under UL485, an outlet can. It only has a mass insertion force of 13.5 pounds, which is designed to be 12.5 pounds of force. By using copper beryllium instead of brass, as most consumer outlets do, there is less overall resistance, as copper is more conductive than brass, further improving the design. ![](https://nicholasjohnson.blog/content/images/2025/07/Plugy3-Short.gif) Pluggy II in action: We test a sample from each batch of contacts we get from our supplier to ensure quality. Next is the physical geometry of Orange Outlets; we designed the receptacle from day one, knowing it had to last 10-plus years in the wild, and tested numerous designs. Using a robot called Pluggy, we could rapidly test the design in ten thousand cycles, seeing the force change over each cycle using a highly accurate force sensor behind the plug on the actuator arm. ![IMG_6642.jpeg](https://nicholasjohnson.blog/content/images/2025/04/IMG_6642.jpeg) Custom Designed Contacts using specific alloys to achieve 10+ year lifespan ![IMG_6793 2.jpeg](https://nicholasjohnson.blog/content/images/2025/04/IMG_6793_2.jpeg) Capable of holding a 8lb weight wthout faling out. Then there is the fact that Orange Outlets are smart, containing a current monitor inside for the energy grade watt-hour meter and networking chipsets to communicate the data. Each outlet has five temperature sensors, each at a critical part. The internal power supply, relays, and contacts have temperature sensors, and humidity and ambient temperature sensors track the external difference between internal and outside temperatures. This ensures that if any temperature exceeds a safe level, the relay will turn off and notify the user charging and the property manager of a possible issue long before any risk of fire. We did this because, over the years, as early adopters of EV charging, we have seen many images of standard outlets wearing out and having stuff fall out of them, as well as burnt-up or melted outlets. By making fundamental decisions, there are no better Outlets on the market for EV charging in commercial settings. You could install dumb outlets or even one of our competitors, who use off-the-shelf outlets for their products and don’t do the added engineering to ensure long-term reliability and safety. It’s your choice whether the risk of a possible electrical fire with standard outlets is worth the upfront savings. At Orange, we won’t ever risk safety for simplicity or making an extra penny. Our exceptional engineers work hard to ensure we create the safest products possible; we don’t want our buildings burning down. ### External Information: [Chesco family says EV charger nearly caught home on fire: What you should knowNathan Simcox, who is a master electrician, said the biggest mistake he often sees is people using adapter plugs or dryer plugs not rated for charging cars.![](https://nicholasjohnson.blog/content/images/icon/favicon.ico)6abc Philadelphia![](https://nicholasjohnson.blog/content/images/thumbnail/14619752_040424-wpvi-investigtion-ev-charger-warning-11pm-CC-vid.jpg)](https://6abc.com/post/chester-county-pa-family-says-ev-charger-nearly-caught-home-on-fire--what-to-know-before-you-buy-a-electric-car/14617760/?ref=nicholasjohnson.blog) [How much does it cost when firefighters put out fires in homes that don’t have homeowner’s insurance in California?Answer (1 of 2): A lot. But fire fighting is a public service provided for by the tax payers. No fire chief in the country would solicit “donations” that if not paid might result in your property burning to the ground because you got put on a “pay no nevermind” list. That’s a Ray Bradbury book.…![](https://nicholasjohnson.blog/content/images/icon/-4-images.favicon-new.ico-26-07ecf7cd341b6919.ico)Quora![](https://nicholasjohnson.blog/content/images/thumbnail/-4-images.share_default_image.png-26-2f12660e125b218f.png)](https://www.quora.com/How-much-does-it-cost-when-firefighters-put-out-fires-in-homes-that-dont-have-homeowners-insurance-in-California?ref=nicholasjohnson.blog) - [Chesco family says EV charger nearly caught home on fire: What you should know](https://6abc.com/post/chester-county-pa-family-says-ev-charger-nearly-caught-home-on-fire--what-to-know-before-you-buy-a-electric-car/14617760/?ref=nicholasjohnson.blog) - [Types of NEMA](https://www.quora.com/How-much-does-it-cost-when-firefighters-put-out-fires-in-homes-that-dont-have-homeowners-insurance-in-California?ref=nicholasjohnson.blog) ### Building EV Hardware To Solve Installation Problems URL: https://nicholasjohnson.blog/why-orange-needs-to-exist/ Last updated: 2026-01-04T01:07:53.000Z I fell into starting an EV charging company, not because I wanted to start a new company, but because I wanted to solve problems that would be monumental to humanity's sustainable future. The story that most people know about Orange is a cheeky anecdote from a podcast and friends. "I lost a bet to Sven Thesen, a Nobel Prize winner, for discovering specific gases directly responsible for climate change while working at PG&E. Sven is a passionate EV enthusiast; anyone who knows him has probably been convinced to borrow one of his EVs to experience what makes them better than your average polluting commuter car. However, this is different from your usual test drive. He lets people take weekend trips to explore what it's like to live with an EV. **So, in 2019, when Sven and Marc asked for help to test an idea to install and monitor level 1 outlets for EV charging in multiple-unit commercial properties, I decided to help them, mainly because I’m a massive EV advocate myself and understand the fundamental benefits of someday having 60-70% of the world's vehicles transition to electric, and two because I greatly respected them.** ![](https://nicholasjohnson.blog/content/images/2025/06/image.png) Each Way was 15.5 Miles and about 20-30 minutes Commuting per Day The so-called bet I lost was that most drivers' daily energy needs did not fill up each night from 0% to 100% state of charge (SOC), but rather a smaller amount of energy to drive the national average of [39 miles per day](https://www.energy.gov/eere/vehicles/articles/fotw-1332-march-4-2024-daily-vehicle-miles-traveled-varies-number-household?ref=nicholasjohnson.blog). To validate this idea, I commuted from Redwood to East San Jose for six months using only a level one outlet. Sven won the bet because, in those six months, I never needed to stop and use public charging during my daily commute to "opportunistically charge". While a level 1 outlet covered all my commuting miles, it was painfully slow, but it worked! This bet made me rethink the problems that need to be solved to expand access to energy for a more sustainable future as electric transportation becomes the new normal. The world is undergoing a significant shift in how transportation and energy systems interact. As we witness accelerated EV adoption, the need for accessible, affordable, and reliable charging infrastructure becomes increasingly urgent. However, in 2020, the charging landscape needed to be addressed, with a few large players focusing on fleets and public charging, leaving most charger sales for single-family homes to local electrical installers. The company was born out of a desire to solve one of the most pressing challenges facing EV owners—access to convenient energy, especially for those living in multifamily properties and workplaces. This paper outlines why Orange needs to exist, focusing on four core problems: the accelerated growth of electric vehicles, integrating transportation with the grid, ensuring energy security, and mitigating environmental impact. # **The Accelerated Growth of Electric Vehicles** ![](https://nicholasjohnson.blog/content/images/2025/04/data-src-image-fabb8122-4f02-4a83-80f6-f632833df484.png) Source: [https://www.iea.org/reports/global-ev-outlook-2024](https://www.iea.org/reports/global-ev-outlook-2024?ref=nicholasjohnson.blog) Orange was founded primarily due to the rapid growth of electric vehicle (EV) adoption. Electric vehicles were adopted more rapidly than many had anticipated, but the charging infrastructure still needs to catch up. In 2020, Orange identified that while single-family homes could easily accommodate EV chargers by tapping into existing electrical systems, multifamily properties faced significant challenges. These properties, home to many urban dwellers, required additional infrastructure to support electric vehicle (EV) charging. The lack of reliable access to charging stations deterred many potential EV owners in these settings from making the switch. EV owners often cite the convenience of waking up to a fully charged car as one of the major perks of owning an electric vehicle. However, this poses a challenge for many residents in apartment complexes, resulting in a clear divide between those who can charge their vehicles at home and those who cannot. With companies like Orange to bridge this gap, the full potential of EV adoption can be realized; residents in multifamily properties will be more inclined to opt for electric vehicles. # **Coupling Transportation with the Electrical Grid** A second, more profound issue driving the need for Orange’s existence is the integration of transportation energy with the national electrical grid. In the past, transportation energy came from refined petroleum products distributed to local gas stations. With the increasing presence of EVs in the market, a new challenge has emerged: the shift towards drawing transportation energy from the electrical grid. This coupling means that, for the first time, electricity used for homes and businesses must also power vehicles, essentially doubling energy demand in numerous scenarios. For instance, in an apartment complex where residents consume a significant amount of energy, adding EV charging further strains the system, often necessitating costly service upgrades. Orange has established itself as a company that solves this problem by optimizing energy management through intelligent charging solutions. Unlike traditional public Level 2 chargers found in multifamily properties, which often consume excessive power beyond drivers’ needs and necessitate costly infrastructure upgrades to meet rising demand, Orange’s innovative approach, particularly its power management systems, allows properties to add more charging stations without requiring these upgrades. By managing when and how EVs charge, Orange maximizes the use of existing energy infrastructure, reducing the need for disruptive and expensive service upgrades. This solution is crucial for scaling EV charging infrastructure at the rate necessary to support mass adoption of electric vehicles. # **The Economic Viability** A third key reason for Orange’s existence is the economic burden that property owners face when installing EVs. This enables property owners to install more chargers without exceeding their budgets, thereby creating charging stations. For example, California has ambitious goals to install 1 million chargers by 2030\. However, the current installation cost, often between $7,000 and $13,000 per parking spot, makes achieving this target difficult. Property owners, particularly in multifamily buildings, are unlikely to invest in charging infrastructure when the costs are so high, and the return on investment is unclear. Orange addresses this by offering a more cost-effective solution. The charging outlets are designed to meet the daily needs of 99% of EV drivers while keeping installation costs as low as $2,000 per parking spot, significantly lower than those of traditional models. This cost reduction enables property owners to install more chargers without exceeding their budgets, thereby making EV charging infrastructure more accessible and scalable. Furthermore, Orange’s model enables property owners to recoup their investment through energy markup, creating a business model that excites rather than discourages them. By aligning the interests of EV owners and property developers, Orange has developed a system that can foster widespread EV adoption. Many programs require governments to invest significant funds in incentivizing EV charging infrastructure in multi-family and disadvantaged communities. The issue lies in historical data showing that the average cost of installing a level 2 AC charger, which used to have a link on the CEC website breaking down the average price of incentive-based installations, ranges from $12,000 to $18,000 per charger. | \# of Parking Spaces | 1,000,000 | | | | | | --------------------------- | ---------------------------------------------- | --------------------------------- | ------------------------ | ---------- | ----------- | | **Options** | **Avg Cost per Parking Space 400 ft Home Run** | **Total Cost For 1000000 Spaces** | **Savings over L2 EVSE** | **$/kWh** | **% Saved** | | L2 Hardwired EVSE (60A) | $ 17,804 | $ 17,804,000,000 | $ (7,524,000,000) | $ 1.55 | \-73.19% | | **L2 Hardwired EVSE (40A)** | **$ 10,280** | **$ 10,280,000,000** | **$ -** | **$ 1.34** | **0.00%** | | NEMA 14-50R | $ 12,012 | $ 12,012,000,000 | $ (1,732,000,000) | $ 1.56 | \-16.85% | | NEMA 6-50R | $ 12,012 | $ 12,012,000,000 | $ (1,732,000,000) | $ 1.56 | \-16.85% | | NEMA 14-30R | $ 9,448 | $ 9,448,000,000 | $ 832,000,000 | $ 1.64 | 8.09% | | **NEMA 6-20R** | **$ 3,128** | **$ 3,128,000,000** | **$ 7,152,000,000** | **$ 0.65** | **69.57%** | | NEMA 5-20R | $ 2,973 | $ 2,973,000,000 | $ 7,307,000,000 | $ 1.24 | 71.08% | The table above outlines the installation costs for a commercial-grade circuit exceeding 400 feet, suitable for either an outlet or an EVSE. As states aim to use ratepayer funds to scale EV charging infrastructure, prioritizing installation costs is crucial and a primary concern. For instance, California has set a goal to install 1 million chargers by 2030, with approximately 230,000 chargers already in operation over the last five years. By opting for a low-powered Level 2 charging solution, the states could save around $ 7.2 billion while targeting one million chargers within the next five years. Orange started by developing a scalable, economically viable charging platform even without incentives. To achieve this goal, we recognized that energy is a tangible commodity, as significant as one of the fundamental physical resources of modern civilization. This means that capital expenditures (CAPEX) emerged as the primary variable controlling the payback period when all other variables and factors were constant. Orange’s innovative solution delivers a remarkable 70% cost savings compared to 40-amp level 2 charging stations. This efficiency enhancement allows residents to connect more vehicles to charging points during working hours and overnight at home. Our approach saves money by eliminating the hassle of constantly moving cars around and the time wasted searching for an available charging station. It generates future value, enhancing the convenience of owning an electric vehicle. # **Energy Security and Sustainability** Beyond addressing the immediate challenges of EV charging, Orange has a broader vision for its role within the energy ecosystem. Rather than simply being a charging provider, the company positions itself as an “access to energy” company.. As the world transitions toward a future powered by renewable energy sources such as solar, storage, and potentially nuclear energy, Orange is dedicated to promoting efficient and sustainable energy utilization by leveraging software within buildings. The company’s dedication to optimizing energy utilization through demand response technology, such as charging vehicles during periods of lower demand and potentially exploring V2X technologies, showcases its commitment to a sustainable and secure energy future. Moreover, Orange acknowledges that while renewable energy sources like solar power offer significant potential, the key challenge lies in effectively delivering this energy to end consumers. By addressing the “last mile” issue in energy delivery, particularly in urban environments with prevalent multifamily properties, Orange is pivotal in facilitating a seamless and practical energy transition for property owners, drivers, and utilities as they strive to adapt to evolving energy needs. # **User Experience with Better Connectivity Architecture** Orange began examining the standards for EV charging in 2020, when many companies used Wi-Fi or cellular connections alone to connect chargers to charging stations and a centralized management system for access control and payments. We quickly encountered situations where chargers didn’t work as they struggled to establish a connection without adding thousands of dollars of networking gear to the garage. Even then, users often didn’t have a cell signal to connect to the cloud and start a charging session. ![](https://nicholasjohnson.blog/content/images/2025/04/data-src-image-74f11288-813d-4edb-9591-bc8e675ad353.png) Underground parking structures and sizeable concrete parking garages quickly proved challenging— Orange set out to solve this problem by rethinking the network architecture of charging. To achieve the desired 99% successful charge session, compared to an industry average of 74%, (Add Source), we built our own IoT network and software architecture that uses a mesh network between chargers to get data and OTA updates while utilizing Bluetooth between users' mobile phones to start and stop charging sessions. In 2024, we achieved a 98.5% success rate across 24,000 unique charging events. We are continually improving this technology by investing in our IoT stack to enhance how chargers store and manage data, authenticate users, and process payments. Unfortunately, to achieve this impressive success rate, we have had to move away from the industry’s poorly architected standards and sometimes don’t qualify for incentives. However, we believe that, in the long run, our decisions will prevail, as they save significant money during installation to network Orange chargers and improve the user experience far beyond what other vendors have achieved. # **The Environmental Impact** We created Orange to expand the market, enabling more people to conveniently own electric vehicles and lead lifestyles that contribute to a sustainable economy. While we recognize the health benefits of electrifying more appliances and transportation to reduce emissions linked to illnesses like asthma and heart disease, we also believe that cleaner air, better energy use, and a resilient grid help slow the effects of emissions contributing to climate change. BloombergNEF has conducted extensive research over the years to understand a net-zero future, and numerous scientists have modeled the effects of emissions on greenhouse gas impact. 1. CO2 Emissions Reduction from EV Adoption (2015-2022): The reduction in carbon dioxide (CO₂) emissions has increased significantly as more EVs hit the road, reaching an estimated 250 million metric tons of CO₂ reduced by 2022. 2. NOx Emissions Reduction from EV Adoption (2015-2022): Nitrogen oxide (NOx) emissions, a significant contributor to air pollution and smog, have been reduced by 600,000 tons by 2022 due to the increasing adoption of electric vehicles (EVs). Switching to electric vehicles has the potential to significantly improve air quality and reduce emissions, particularly in urban areas with high pollution levels, which often disproportionately affect marginalized communities. Governments encourage this shift through financial incentives, infrastructure developments, and regulatory measures. With the ongoing rise in adoption and the transition of energy grids toward renewable sources, the long-term environmental and public health benefits will be considerable for all individuals as electric vehicles become more widespread and accessible, thereby reducing energy costs. # **Conclusion** Orange exists to address the pressing need for scalable, cost-effective, and efficient EV charging infrastructure, particularly in multifamily residential settings. Our solutions effectively address the logistical challenges of installing and maintaining charging stations and providing property owners with economically viable options. By optimizing energy use and reducing the need for costly infrastructure upgrades, Orange paves the way for widespread EV adoption without burdening the electrical grid. Companies like Orange play a pivotal role in ensuring that infrastructure development aligns with technological advancements as the global community transitions towards a more sustainable energy future. ### Why Orange Had to Exist: The Origin Story URL: https://nicholasjohnson.blog/why-orange-had-to-exist-the-origin-story/ Last updated: 2025-12-22T06:31:39.000Z In 2020, I wasn’t trying to be a founder. I just wanted to solve a problem that was slowing the adoption of electric vehicles, and what was on the market wasn't scalable. At the time, I was renting an apartment in a building that lacked EV charging facilities. I had worked on the Model 3 at Tesla and spent the last four years founding Lyt.ai, an AI traffic management company. While working on fundraising for Lyt, I became friends with Seve Thesen, Marc Geller, and Marc Taprneing, who all saw that the problem of getting access to energy for residents of multi-unit properties was almost impossibly complex and expensive. For people living in single-family homes, adding a simple charger to the garage was easy enough, though not always economical. However, this access to overnight charging made owning an EV better than owning any other vehicle. Property owners didn’t want to pay for installation. Tenants didn’t always have access to dedicated spots. Utilities weren’t helping with expensive and time-consuming service upgrades. Every solution we saw was built for the wrong customer: commercial fleets, single-family homes, or public parking lots. No one was solving for where most Americans actually live—multi-family housing. So, we started Orange to scale EV charging to as many parking spaces as possible per dollar spent. The vision was simple: make EV charging as easy and scalable as Wi-Fi. Something you could plug into your wall. Something tenants could activate and pay for. Something that didn’t require $100,000 in electrical upgrades. We didn’t want to build another charger—we tried to re-architect how charging infrastructure was deployed, managed, and monetized. What I saw in 2020 was a market shaped by compromise. - Chargers that required utility upgrades. - Networks that broke under scale. - Installations that made no financial ROI for landlords. Orange needed to exist because: 1. **The existing model was broken.** It was top-heavy, too slow, and far too expensive. 2. **EV adoption was going to outpace infrastructure.** Especially in cities and dense housing markets. 3. **Property owners needed a business model.** Not a charty project. We took a fundamentally different approach: - Chargers that work on existing power constraints. - Plug-and-play architecture. - A mesh network that doesn’t rely on Wi-Fi or cellular in underground parking. - A mobile-first user experience for tenants. - A dashboard for landlords that shows ROI—not kilowatts. - A system that didn't break and requires ongoing maintenance. This wasn’t just a hardware company. We were building a distributed energy platform to accelerate access to energy. We developed Orange Outlet, a compact, affordable, power-efficient, and maintenance-free solution that enables property owners to provide access to energy in a scalable and economical manner. The ecomics mattered, so we developed a business model for property owners and managers that would transform a money pit into a revenue-generating asset, attracting high-value residents to their properties. [Public Orange Enabled Property Onwer Business Model![](https://nicholasjohnson.blog/content/images/icon/spreadsheets_2023q4.ico)Google Docs![](https://nicholasjohnson.blog/content/images/thumbnail/AHkbwyLHEEBANsnYqs4gefDCpIh3VIjgapytcR8tmBiNoOcOJ51pj61Qy7wJsONwlHD6ipADuucObS7OFQklbYVkG3Puxdcfm8jKRmM-qmpT-2DjJcBtYkeT-w1200-h630-p)](https://docs.google.com/spreadsheets/d/1hKmubsTegJugSCv9-018RVGl-1Xr2MjTEBgPO-PC7S0/edit?usp=sharing&ref=nicholasjohnson.blog) This simple business model turned EV charging from a loss leader to an ROI that was somewhat more attractive to property owners. This was just the beginning of the journey as we started developing. With states like California working on new building codes that could require 50% of parking to be electrified, this mode is critical to ensure property developers can still build new developments in many cases. Let's take a moment to consider an example of how our solution solved these issues. Property Example details, - Number of units - 100 - Total parking space 130 - Average Rent $2000 Based on the above details, under the California proposed CalGreen codes, changes of 50% this property woudl require 50 chargers. | Power | Outlet | L2 EVSE | | -------------------- | ------- | ------- | | Amperage Per Circuit | 20 A | 40 A | | Power Per Circuit | 3.8kW | 6.6kW | | Voltage | 208/240 | 208/240 | This means for 50 chargers, the property would need an additional 330kW of power budgeted. This easily pushes developments into adding a new transformer and service with the utility, which can cost $100k+ and add delays that cost even more. Compare this 330kW to Orange Outlets' 165kW, and you end up with half the power budget, often able to avoid adding a separate utility connection just for EV charging. | | Standard L2 Charging | Orange Outlet | | ---------------------------------- | -------------------- | ------------- | | Hardware Cost | $2,500 | $750 | | Electrical Permits & Licenses Fees | $500 | $500 | | Software Commissioning | $550 | $0 | | Networking Geer | $1,000 | $0 | | Project Management Fee | $0 | $0 | | Average Installation Cost | $3,500 | $1,500 | | Total Fixed Install Cost | $8,050 | $2,750 | | Charger Annual Service Fees | $320 | NONE | | Commissions on Revenue | 10% | NONE | With a highly technical team, we developed the hardware and software from the ground up, thereby avoiding complex installation and commissioning processes. Think of it as the Apple of EV charging. Each outlet features Wi-Fi, cellular, and Bluetooth capabilities, paired with a proprietary edge-computing-based software architecture rather than a cloud-based one. This means that the chargers work even if the signal is lost for several days. With the Orange app installed and an account set up, a driver can even start a charging session even if they don't have connectivity, a common issue with other chargers in underground parking structures often found in significant multi-family developments. This build of networking into the charger and software architecture has another benefit. The property doesn't need to add Wi-Fi to the existing parking structure, saving not only the added hardware cost but also the installation and IT costs associated with setting it up. Most OPCC chargers require commissioning, which takes about 20 minutes per charger and requires a good network connection to work. Assume $150/hour for a low-voltage electrician to complete this task, and those 50 charges take 16 hours just to set up after installation, adding $ 2,400 in additional cost. Unlike most EV providers that use OCPP activation standards, Orange deviated from the norm to deliver an exceptional user experience that saves on installation costs and provides better connectivity through Bluetooth between the phone and charger, ensuring seamless operation. Orange was born from a deeply technical and deeply personal problem. I didn’t want another startup—I wanted to build the infrastructure I wished existed. Orange had to exist. Because the future was already here, it just didn't have enough charging stations to plug in and charge effortlessly! --- Stay tuned for more posts about how we scaled, what we learned, and what comes next. If you’re building in climate or energy—reach out. We need more people thinking differently.