With the door closed, China found a window. It’s been over a year since Europe introduced tariffs on Chinese-made EVs. The Economist published an article that is blunt in its conclusion - this policy failed. So much is clear by looking at the volume of Chinese car exports to Europe over the last year—a dip, followed by a rise, not only in hybrids, but in EVs too. I wonder what figures European automotive lobbyists used when trying to convince the European Commission that Europeans don’t want to buy EVs? It seems that instead of revving up factories and seizing the advantage of the temporary dip in Chinese EV exports, EU automakers decided to spend more time in Brussels. Can’t blame them - cafes are much nicer than factory floors.

No one likes Brussels. But we should. The news of the week is the European Commission's decision to ease 2035 emission targets for European car-makers. Funny enough, Brussels managed to offend both ends of the automotive opinion spectrum - those who believe that the European automotive players should switch to electric vehicles ASAP, and those who push for extending the sales of internal combustion engines. At the beginning of the week, right after the announcement of the easing of the ban, my LinkedIn was flooded with laments for Europe. People were calling it a huge setback and a death sentence for the European auto industry. Then, towards the end of the week, a different chorus of voices rose, this time from the automotive giants themselves, who also blamed the new regulations for increasing costs and deterring investment! “This package doesn’t do the job”, said Antonio Filosa of Stellantis, “disastrous”, echoed Hildegard Muller, president of VDA. So what’s the deal here? First, OEMs can continue selling ICE cars and release 10% of their 2021 CO2 by 2035, instead of going to zero. A clear blow to EV transition, and a win for ICE. But to comply with the 10% requirement, emissions must be offset with low-carbon steel and sustainable fuels. Which, in the words of Matthias Schmidt, an industry analyst, would make European ICE cars “haute couture Swiss watches of the motor industry”. When I worked as a lobbyist, I saw how immense industry pressure could be on the rank-and-file lawmakers. You have many who are paid far more than you and have access to all the consultants and resources you’ll never have, trying to convince you that their point is correct. You do your own research, but you are understaffed and out of time. Then, there is a political dimension too. Your bosses manoeuvre for power, and they can cut deals in the backrooms that will nullify your effort. So the Brussels decision seemed to be a gambit - sacrifice some CO2 targets, but make exploiting this option too expensive. On the face of it, the ICE industry wins. In reality, they’ve been duped, as they now realise. Hope that after spending so much firepower on pushing through a deal that ultimately hurts them, they’ll go back and double down on EVs. Especially since Brussels have announced the same week a carrot of €1.8B for the battery industry.

I love my Mac. It's seamless integration, the way everything just works, the illusion of perfection. But every once in a while, my Mac freezes. The spinning wheel appears. A polite message informs me that “the system has been unexpectedly shut down”. So every time I hear someone talk about a software-defined vehicle, my brain does this: I’m driving down a German autobahn at 250 km/h, in the left lane. Everything humming, and then the dashboard lights up: The system has unexpectedly crashed. Reboot in progress. That’s usually where I wake up, drenched in sweat. Software-defined vehicles (SDV) are presented as inevitable. As if resisting them is like resisting gravity or USB-C. So I dug deeper: • What’s actually behind SDVs? • Does anyone make real money from them? • And, most importantly, do users even want them? Here’s what I found. First: SDVs are not cheaper to produce Despite what some decks suggest, there is no solid evidence that SDVs reduce manufacturing cost. Yes, they reduce the kilometers of wiring.
Yes, you may consolidate the electronic control unit. But then: • you add high-performance compute • expensive chips • redundant safety architectures • massive software teams • cybersecurity • OTA infrastructure • compliance overhead. Any wiring savings are quietly eaten by silicon and software. Now add one more thought experiment: remember the last global chip shortage? What happens next time, when all OEMs depend on central compute to ship cars? Exactly. Second: SDVs do reduce lifecycle costs This part is real. Early fault detection, OTA fixes instead of recalls, better post-sale service management.
Fewer angry dealer visits. From an OEM perspective, this is meaningful. But let’s be honest about who really loves this: 
👉 Insurance companies. - Better data.
 - Better risk models.
 - Better pricing.
 - Lower uncertainty. Sleep well, actuaries! SDV has your back. Third: customers don’t want “digital cars” They want: • navigation that works, • charging that doesn’t lie, • an app that doesn’t log them out, • software that doesn’t feel like a beta test. And they want all of this… for free. The audacity! Why won’t customers pay subscriptions so Volkswagen can live like Netflix? Why won’t they understand that heated seats are a content strategy? Because customers already have:
- Netflix.
- Spotify.
- Apple Music.
- Cloud storage.
- Five forgotten app subscriptions. They are not emotionally ready for Car+. I’m not ready, that’s for sure! Who makes money from SDVs? Who actually makes money from SDVs and their fashionable cousin, connected services? It is not the consumers, not the OEMs (at least not directly), and not the app stores in cars. The only group consistently smiling are 👉 Commercial fleets. Fleets care about: • uptime, • total cost of ownership, • predictive maintenance, • compliance, • risk reduction. They pay.
They renew.
They don’t complain about UX animations. Boring. Rational. Profitable. The Luddite conclusion SDVs are not a gold mine. They are more like an operating system upgrade for OEMs: • reduce operational pain, • shift cost curves, • improve control, • make recalls less terrifying. But they do not magically unlock subscription riches. And they certainly don’t make cars cheaper. So yes, SDVs are probably inevitable. Just don’t tell me they’re cheap, customer-loved, or a money-printing machine. And please…
before we ship them at scale… Let’s make sure the reboot message never appears at 250 km/h. Don’t ask me how.

There’s an old anecdote I keep returning to. The rabbits in the forest are tired of being hunted by foxes and wolves. They gather the courage to approach the wisest creature they know—the owl—and ask for help. The owl thinks for a moment and says: “You have to become hedgehogs.” The rabbits look puzzled. “But how?” And the owl replies: “I’m a strategist, not a tactician.” About five years ago, BP hired McKinsey to help steer the company away from oil and into renewables (you know where it goes). The effort lasted four years—exactly as long as the CEO who initiated it. Shell made similar moves. Both transformations collapsed the moment leadership changed. Yesterday, I was reading the FT’s long read on BP and Shell’s failed attempts to reinvent themselves as clean-energy companies, and thinking, is it ever possible for oil and gas companies to pivot to clean energy, or are these efforts doomed to fail? A quote from one executive captures the key difficulty: “I don’t get it. I don’t know why we’re investing in this—the returns are so low.” BP and Shell tried to become electricity companies. But utilities operate on completely different economics: lower margins, longer, slower cycles, and value creation that takes a decade or more to materialise. It took Shell’s LNG business 10–15 years to become its crown jewel. No one seems to be prepared to wait that long for renewables. And no amount of consulting work—no matter how expensive—could compress that journey into a single CEO’s tenure. You can tell the rabbits to become hedgehogs, but maybe your DNA just won’t let you grow the spines.

2025 was the year FOAK reality finally caught up with FOAK ambition. After a decade of increasingly bullish climate-tech projections, the past twelve months showed what it really takes to deploy first-of-a-kind projects, and what you lose when things don’t work. The data reflects this: 42 FOAK projects were scrapped in the US in 2025, up from just 14 in all of 2024—a threefold increase in a single year. 69% of investors now expect FOAK funding to shrink through 2026, and most believe these projects will face the steepest financing conditions in years. There is no such aggregated data for the EU or the rest of the world, unfortunately. But this same period also delivered some of the strongest FOAK wins we’ve seen to date. The picture is not one of collapse, but one of sorting. Technologies that are ready are scaling. Technologies that are not ready are getting exposed. 
This is my first annual review of climate FOAK projects. I’ll focus on the three successful that struck me most, the three unsuccessful that teach valuable lessons, and I’ll end with three FOAK projects to watch in 2026. What FOAK Succeeded in 2025 1. Kudgi Liquid CO₂ Energy Storage (India) – “Success by Leaving the West” CTVC highlights Kudgi as one of the few actual FOAK wins of 2025, and the reasons are instructive. Instead of fighting for permits, incentives, and EPC labour in the US or EU, Energy Dome licensed its technology to a local EPC in India. Here’s what they got for such geographic pivot: - Build time under 3 years - 1-year construction phase - CAPEX of $205/kWh vs $367/kWh in Europe and $500/kWh in the US Even if you target the US and EU markets, your FOAK need not be there if it doesn’t make economic sense now. FOAK is the most expensive part of your journey, so if you can avoid compromising technology or market, then shifting to a cheaper geography may be the difference between success and failure. 2. Fervo Energy – Cape Station Geothermal (Utah, USA) While not yet a full FOAK execution success, Fervo continued its execution streak, raising $206M in June 2025 to build the first 500 MW enhanced-geothermal FOAK in North America, with the first 100 MW expected online in 2026. Geothermal has been sipping through many media outlets to me, but this year, I’ve managed to avoid writing any opinion posts on it. Looks like next year will be different. 3. Brevik CCS & Northern Lights (Norway) Commissioned in June 2025, this could be one of the decade’s most important industrial decarbonization FOAKs: - 50% emissions reduction at Norcem’s cement plant - $3B total cost, where $2B was public money - 5-year build time It worked, but only because the state carried two-thirds of the financial load. And its offtake model is book-and-claim: the cement itself is still produced conventionally; the CO₂ reduction is “claimed” via accounting. A functional FOAK, but not a replicable one yet. After all, no NOAK can expect 2/3 of funds from the government. What Failed in 2025 If successes teach us how to build FOAKs, the failures teach us how to avoid burning a hundred million dollars before discovering the physics doesn’t scale. 1. Climeworks – Mammoth DAC Under-Performance The Mammoth plant captured only 105 tons in its first ten months, against a design capacity of 36,000 tons per year. The company failed to capture carbon at any meaningful scale. I’ve written about Climeworks' modular approach to FOAK, showcasing it as one of the ways to make your FOAK a success. Well, it certainly seems that if your underlying tech isn’t working, then even building modular won’t save you. This is a textbook FOAK outcome: the plant taught the company what the tech actually is. 2. Natron Energy – Sodium-Ion Collapse After raising $363M over 13 years and opening its first US sodium-ion facility in 2024, Natron shut down in September 2025. Sodium was (and still is) touted by some as an LFP/NMC killer. While I am not with this cheering crowd, I do see use cases for sodium. Natron’s case just shows that even with good technology, juggling many things during FOAK can really mess things up. 3. Air Products – Mass Cancellations and a $3B Write-Off 2025 was the year FOAK hydrogen went from “hyped” to “paused.” Air Products demonstrated it best, when it exited three US-based hydrogen projects in February 2025 and recorded a write-off exceeding $3B. When one of the largest industrial gas companies in the world walks away from its own flagship FOAKs, it signals a deeper structural problem - like hydrogen being a really poor energy carrier. And that’s on top of: - No bankable demand for premium-priced hydrogen - Unproven economics for multi-gigawatt electrolyzers - Rising cost of capital hitting megaprojects hardest Three Projects to Watch in 2026 The projects below aren’t guaranteed successes. What makes them worth watching is precisely the opposite: they will give us the clearest signals about which parts of climate tech are ready for true scaling. 1. Lyten – The Northvolt Takeover If Lyten’s 3D-graphene-based lithium-sulfur batteries take root inside Northvolt’s assets, 2026 could see the most dramatic pivot in the battery sector since LFP’s global takeover. This is FOAK at the intersection of distressed assets and breakthrough chemistry. I’ll be watching this one closely! 2. H2 Green Steel – Boden Facility Europe’s first large-scale and long-hyped hydrogen DRI plant will test: - Whether green hydrogen can run an industrial process reliably Whether customers actually pay green premiums Whether steel decarbonization will be built on hydrogen My money is on H2 Green Steel going the way of Northvolt, but for the reason of initially flawed hydrogen economics, rather than poor execution. But you never know, and we could have both! 3. Form Energy – Multi-Day Iron-Air Battery The FOAK 10 MW / 1000 MWh system planned for first operations will answer a decade-old question: Can long-duration storage be built without lithium, at grid scale, at a cost utilities will accept? If the answer is even “partially yes,” then the case for overbuilding renewables will be even stronger. Otherwise (and highly likely), gas turbines will prove to be the only viable alternative to cover peak loads or renewable shortages. Looking Ahead The pattern emerging from 2025 is clear: - FOAKs succeed when and where they are buildable, not when they are perfect. - FOAKs fail when physics or economics break at scale, not when founders lack vision. - Public capital still carries the heaviest industrial FOAKs, and that isn’t likely to change in 2026. - Investors are pulling back, but the best technologies are moving forward regardless. From January 2026, I’ll be monitoring FOAK projects around the world on a monthly basis and will dedicate one monthly edition of this newsletter to overview of what’s going on in the climate FOAK world. 2026 will bring its own real-world test cases. And—like this past year—they will rewrite strategies more than any policy paper or conference keynote ever could.

#China  is at 50%  #EV  penetration today and on track for ~80% by 2030.  #Europe  sits at 26% and is moving steadily.  The  #US  has just revised its forecast from 40% → 18% by 2030.  That’s the part of the  🎙Redefining Energy  Energy podcast episode on batteries and EVs that stayed with me — the sheer divergence of industrial strategies. China treats EVs as a national industrial priority. Europe frames them through regulation. The US is now pulling back on incentives and standards. Same technology, three trajectories. If you work in  #batteries , #mobility , or industrial policy, this episode is worth listening to. It raises uncomfortable but essential questions about competitiveness, supply chains, and the cost curves we’re all implicitly betting on. 🎧 Have a listen, and tell me what you think:  Is this divergence temporary or permanent?

Public funding is meant to accelerate FOAK projects. In Europe, it often slows them down.  The Financial Times recently ran an article on the EU funding process for climate scaleups. Here is what you can expect if you plan to get public funds for your FOAK:  - 3000 hours of paperwork. - Around €85,000 cost per application - 20% chance of success and  - 6% chance of actually getting any money  For FOAK projects, this is an existential issue.  When you’re building the first commercial plant, your schedule is your financing.  Your suppliers lock in prices for 60–90 days.  Your EPC window keeps closing.  Your team burns payroll while waiting for a letter from Brussels.  And the moment your budget moves… you’re back to square one in the eligibility checklist.  This is why so many European hardware founders quietly do the thing that  Vianode  did: raise private capital first, start building, and treat grants as a maybe-later bonus. It shouldn’t be this way. Europe needs more FOAK factories — not more abandoned applications. But until funding matches FOAK reality (speed, clarity, and decision-making within the lifetime of a CAPEX quote), founders will keep choosing survival over subsidies.  If you’re a climate-tech founder navigating this maze — or deciding whether public money fits into your FOAK timeline — I'd be happy to compare notes!

This week’s news about Russian Porsche owners having their cars remotely deactivated stayed with me longer than I expected. One moment, you have a €150,000 sports car. The next, you have a very heavy, very pretty paperweight—because someone, somewhere, pressed a button. It is a neat illustration of the direction we’re drifting toward with “connected” and “software-defined” cars. In my recent article on robotaxis, I wrote that we’re stepping into a world where the economics make little sense, but the technological momentum is unstoppable. This Porsche episode is from the same story—just on the ownership side of the equation. We used to have a clear divide. You owned your car, phone, TV. The state owned (or at least regulated) the roads, the networks, the infrastructure. Now, the lines blur. You buy the hardware, but the “soul” of the product—the software, the control surface, the kill switch—belongs to someone else. And that someone often sits in a corporate office with little accountability and even less transparency. With cloud-based controls, you don’t really own your car anymore. Porsche just demonstrated that—even if you pay top dollar, your property rights are conditional on the goodwill of a faraway server. This isn’t just an annoyance. It chips away at the foundations of any healthy economy. Property rights were supposed to be simple: you buy a thing, it’s yours, and no algorithm can take it away. Combine this erosion with the “winner-takes-all” logic of modern tech platforms, and you get a world where even wealth doesn’t buy independence. You’re still tethered to a digital leash owned by a megacorp—and you hope the person holding the leash had a good morning and a stable internet connection. Robotaxis show us what happens when mobility becomes fully automated but economically fragile. The Porsche case shows us what happens when ownership becomes fully digital but legally fragile. Two sides of the same shift. And both raise the same question: What does “ownership” mean in an automated, connected world—when the “wires” run not to your garage, but to someone else’s cloud?

Europe wants to learn how to make batteries. What it gets instead is assembly lines. Last week, I stumbled on a report I somehow missed when it came out — the excellent T&E-commissioned study by Carbone4 on foreign battery investments in Europe. It’s one of those documents where you read ten pages, stare at the ceiling for a moment, and mutter: So it’s not just me seeing this. The picture they paint — and the picture in the infographic below — is surprisingly consistent with my own experience trying to localise wind turbine manufacturing years ago. Europe keeps talking about cooperation, know-how, and technology partnerships. What it is actually getting is… well, immovable property. We talk a lot about localisation. But what gets localised? In the ten years I’ve spent around manufacturing — wind turbines, CHP units, battery electrodes — I’ve met many people who say they localise technologies. But I don’t know anyone in Europe who has actually done it with a foreign partner in the full sense: design, IP, core processes, equipment, operations — the whole stack. Factories, yes — Europe has plenty of those. But true localisation? Real transfer of know-how? Tell me if you know. I don’t. And batteries are heading the same way. A growing chorus in Europe now argues that we should “cooperate more with China” because Europe “lacks battery know-how.” On LinkedIn, every announcement about CATL in Spain or Gotion in Slovenia is met with applause emojis and optimism about Europe “catching up.” I’m all for cooperation. But cooperation without technology transfer is not an industrial strategy - it’s outsourcing. And the report confirms that is exactly what’s happening. My wind turbine déjà vu When I was tasked with reaching >60% localisation of wind turbine manufacturing in Russia, I spoke to everyone who would care — Goldwind, MingYang, Siemens, GE, Vestas. At every interaction, I had a list of components for localisation, and I pushed for it. In my conversations with Chinese manufacturers, the main message was “We will ship you all the critical components, like generators. You can localise towers and some castings.” The negotiation basically ended there. Talking with European manufacturers was not much different. They were open to deeper localisation but on their terms. Location? They decide. Component scope? They decide. Skills transfer? Vague promises and PowerPoints. This is exactly what I now see playing out in Europe’s battery industry. The T&E / Carbone4 findings are blunt: battery technology transfer in Europe is simply not happening The report analysed the two major Chinese–EU partnerships — VW-Gotion and Stellantis–CATL—and the Gotion–Inobat JV. The conclusion is not ambiguous: there is no meaningful transfer of knowledge, IP, skills, or manufacturing expertise. Here are a few highlights. First, the ownership is lopsided. In any partnership, there are those who make decisions and those who follow. If you try to make anything in China, you will have to accept a smaller equity share and strict IP and know-how-sharing requirements. Gotion controls 80% of its JV with Inobat. CATL keeps all core technology in China. VW may own 26.47% of Gotion, but operationally, it's more of a customer than a partner. In none of the EU cases does the European partner hold decisive control over product, process, or equipment. Second, IP transfer provisions are either “limited” or non-existent. Stellantis’ €300M in Spanish state aid came with zero requirements for technology transfer. VW’s partnership secures battery supply, not battery know-how. Third, local supply chain development is not required. Plants rely on imported components — exactly what China wants: export of semi-knock-down kits assembled in Europe. Fourth, fuzzy provisions for local workforce development. To understand how this will play out in practice, look to existing Asian gigafactories in Hungary. CATL and Korean plants rely heavily on temporary migrant labour with high churn, and the top managerial positions are filled by expats from Korea and China, meaning the skills built do not stay in Europe. This is not cooperation. This is exploitation. Europe currently gives foreign manufacturers hundreds of millions in subsidies — €900M to CATL and LG alone — while attaching no requirements for technology transfer, local content, skills development, or even basic environmental safeguards. Europe has leverage — but is shy to use it Here’s the most astonishing part of this story. When I negotiated localisation in Russia, the only thing that helped me was the existence of strict local content rules. The market wasn’t large, but the rules were clear. Most players openly scoffed at the minuscule market size, saying it is not worth their time to even consider localisation there. Europe today is in a stronger position than Russia ever was: Europe is the world’s second-largest EV market. China is suffering from battery overcapacity. The US is largely closed to Chinese suppliers. Chinese companies need Europe. Yet European negotiators behave as if they were the weaker party. True, the EU doesn’t have any local content requirements, but their market size alone should be a big enough bargaining chip. Instead of demanding knowledge transfer, Europe funds knock-down assembly. Instead of making the market conditional — Europe hands out subsidies unconditionally. Instead of requiring local hiring — Europe receives temporary migrant workers with no long-term skills retention. On a side note, Northvolt's case is super interesting, as they also employed many migrant workers, who acquired critical skills, but were basically kicked out by the Swedish government after Northvolt’s collapse. That’s undoubtedly the best example of how to keep crucial, hard-to-develop skills in Europe. What should Europe do? The answers are not complicated. As the report puts it plainly, Europe risks becoming “an assembly plant, not a battery powerhouse.” The checklist Europe needs is something that has no doubt come about many times in the boardrooms and national assemblies: Require majority local ownership in JVs (>51%) Condition state aid on IP transfer and local content Demand training programmes and local workforce targets Use tariffs as leverage for onshoring Enforce grid-based CO₂ limits that favour local production Implement “Buy Europe” rules in procurement Define clearly what is “Made in EU” These measures are nice to have, but not critical. What is sorely lacking is an understanding that the world is changing, a strategic vision for one’s industry and some backbone in negotiations. These alone could score Europeans better deals. There is a caveat, however. Freedom and independence come at a price. You can’t get both technology and cheap batteries at the same time. The European car and battery industry is so squeezed that it values next-quarter earnings above long-term survival. So it sells itself on a low price, securing a battery supply in exchange for future technology independence. When we finally decided on a technological partner in wind turbine manufacturing, we went with a small Dutch startup, not GE or Goldwind, precisely because we would get the technology and know-how we needed. We knew our turbine would be more expensive. We knew we would make many mistakes as we learned to scale up turbine manufacturing. We knew that we would have to design and fund training programs for local workers (we even created a VR exam on generator assembly). Next time you hear about a “European–Asian partnership”… be wary, because: A factory is not a strategy. Assembly is not localisation. And a partnership without knowledge transfer is not a partnership — it is dependency. Europe still has a chance to avoid becoming the world’s green-tech workshop. But it needs industrial policy with teeth, and businesses willing to take a long-term view and make hard bets. Until then, every new battery “joint venture” should be read with caution.

This week started with a review of five pitch decks — from sustainable fishing to fashion, fuels, energy trading and carbon reporting: different industries, different stages, different teams. So, who actually has a chance to survive long enough to build a FOAK? I run them through my 4—step FOAK strategy assessment framework. It’s a process to force clarity on climate impact, market fit, FOAK delivery, and team readiness. It helps to turn scattered assumptions into an actual long-term strategy. Strategy is a living document. Revisit it while you build, challenge your old assumptions, and notice what you actually got right. Here’s the short FOAK strategy checklist I use: 1. Climate impact: Can you eliminate ≥500 Mt CO₂e/year at scale? Are there any rebound risks that your tech would actually do more harm than good? 2. Market: Is there a real market today, not in 10 years? What’s your value beyond CO₂ reduction? 3. FOAK delivery: Manufacturing feasibility, supply chain readiness, cash flow, financing and timeline. And what will you do post-FOAK? 4. Team: Founder–FOAK fit, leadership gaps, and a talent map for who you must hire to run FOAK. The exercise gives you two things every founder needs: - A clear view of your real position in the climate tech system — your impact, risks and capabilities. - A map for FOAK and beyond — milestones, bottlenecks, and the “known unknowns” you’ll meet while putting out fires on the way to your first plant. Once you know your limitations and your path, you can finally plan how to execute, finance and operate your FOAK, and don’t lose the forest for the trees. If you want to pressure-test your own FOAK strategy, I'd be happy to take a look!

Construction sites and factory floors are places I’m far more used to than a laboratory. But last week I was in the Netherlands, meeting a client and visiting two universities to see how their new chemical product is getting ready for its demo project. And I have to say: I haven’t yet seen this level of preparation — on both the R&D and project sides. Every assumption challenged, every risk logged, every experiment linked to a downstream engineering decision. It’s the sort of discipline that quietly tells you: this FOAK has a real chance. Sometimes I hear that you start planning your FOAK when you hire EPCs or sign for a plot of land. In reality, your FOAK begins in the lab — long before the first layout drawing. Science gives you the spark, the “what if.” But execution is what turns that spark into hardware, contracts, and something that runs at nameplate for more than a week. Last week’s trip made me quite confident in my client’s success. The science is solid. But more importantly, the execution mindset is already there. And that’s what scales.

A few weeks ago, I posted a clip from Total Recall where Arnold Schwarzenegger tears the robotic taxi driver out of his seat and takes over manually. For all the futuristic imagery, that scene captured something intimately familiar: you don’t trust a robot. This week, I read two Economist articles back-to-back—one on the US robotaxi sector, one on the Chinese one. And after going through both, one conclusion is impossible to ignore: robotaxi economics still don’t work. Not in the US. Not in China. But both countries push ahead anyway—just for very different reasons. Source: The Economist The US, where unit economics are a tragedy, but TAM slides look magnificent. Start with the US. Here, the unit economics are a tragedy, but the TAM slide looks magnificent. The Economist article on the US essentially boils down to three problems. First, the cars are too expensive. A full US-spec robotaxi stack (lidar–radar–cameras–compute) is a six-figure machine, however you turn it. And these vehicles depreciate fast. Second, the operating costs are comically high. Robotaxis require large teams to monitor fleets, endless software updates, high-bandwidth connections, map maintenance, remote “intervention” staff, and high insurance premiums. Robotaxi companies do replace drivers - with 300 software engineers and 24/7 remote support. Third, regulation keeps the fleets tiny. Robotaxis are allowed to roam only in designated parts of towns, if they are allowed at all. Small fleets won’t let you have economies of scale, which means costs are still high and startups are still unprofitable. Yet the US keeps pushing. Why? Because America still believes every mobility problem can be solved by a startup, a valuation model, and maybe a positive EBITDA slide. The tech optimism is deeply baked into the culture. After all, this is the country that turned coworking into a high-conviction asset class. China, where unit economics still don’t work but the system is trying to brute-force them into working. Now let’s move over 10,000 km west of San Francisco. In China, the robotaxi economics also don’t work, but the system is built to brute-force it until it does. The second Economist article is even more interesting, because despite the hype around China “pulling ahead,” the underlying economics are still shaky - China’s robotaxis lose money too. Even though Chinese robotaxis have one zero fewer on the price tag, per-kilometre costs are still higher than ride-hailing with human drivers. China has lower manufacturing costs and has deployed many more robotaxis, but it's still not enough to be profitable. So why does China look like it’s pulling ahead? Because China’s system solves the problem the opposite way around: if the cost curve doesn’t work, build the scale first and push the cost curve down later. This is the same logic behind EVs, solar, and batteries. Economics today matter less than strategic positioning tomorrow. Despite both countries losing money on robotaxis today, the probability of getting to profitability is much higher in China. First, the hardware is cheaper, and it is getting cheaper. China produces lidar, sensors, compute hardware, and EVs domestically at far lower cost. A US robotaxi is a science experiment on wheels. A Chinese robotaxi is an industrial product rolling off an existing EV supply chain. When your bill of materials is cheaper, your break-even point suddenly looks less like a fever dream. Second, Chinese cities welcome robotaxis and actively redesign infrastructure to support them, with the curious exception of Beijing, Shanghai and some other large cities. In the US every city requires separate approvals, hearings, pilot phases, and local negotiations. In China municipalities compete to be early adopters. Third, scale is a policy decision, not a market outcome. China can mandate large fleets, provide regulatory clarity and subsidise deployments. If robotaxis need 30–50k units on the road to hit favourable economics, China can make that happen. The US cannot—not without a decade of hearings and lawsuits. So why push so hard if the economics are broken? Because in both countries, non-economic drivers dominate. In the US founders need the “inevitable autonomous future” narrative to stay alive. Investors need one last big technological frontier. Companies like Tesla and Waymo need autonomy to anchor long-term valuations. In China robotaxis are part of the national industrial strategy. They support domestic supply chains for EVs, computing, and sensors. They help cities hit local digitalisation and emissions targets. The government wants to define global standards for autonomous mobility. The incentives are political, strategic, and reputational—not financial. A long drive to profitability Robotaxis today resemble green hydrogen was five years ago: lots of headlines, small deployments, great PowerPoints, and a cost curve that stubbornly refuses to bend. But while the US treats robotaxis like a startup race, China treats them like infrastructure. And that difference will define the next decade. The US will probably lead in core autonomy algorithms and chips. China will lead in deployment volume, cost reduction, and profitability. The first positive unit economics will appear in China, not in Silicon Valley, because ultimately, this is an industrial game. And in industrial games, China tends to play the long, patient, scale-first strategy—while the US plays the venture-funded, demo-first strategy. Robotaxis don’t make money today. Unlike hydrogen, there are no physics telling us that they won’t make money tomorrow. With enough scale and cheaper computing power, they will. What will it take? Two things. First, a ton of innovation aimed at improving the efficiency of algorithms and energy use. The hardware has to get cheaper, and it will. When cell-to-chassis finally comes about, that will be one sign. Second, it will need scale, and if anyone can force them to make money tomorrow, it’s the country that already forced EVs, solar panels, and battery factories onto workable cost curves. ---- Original articles, no paywall: https://economist.com/who-will-win-the-trillion-dollar-robotaxi-race?giftId=Y2Y5ZDg0MTQtZmFjZC00OGJhLTk3MDUtOTQ4YWIwYzkxM2I2&utm_campaign=gifted_article   https://economist.com/business/2025/11/26/why-china-is-pulling-ahead-in-the-robotaxi-race?giftId=NzBjYzI0NzMtM2FhOS00NDM4LWE1MDktZTQzZDdjM2ExZWY3&utm_campaign=gifted_article