The Learning Curve You Cannot Buy
What the Western battery effort got wrong, and what the wreckage left behind
Fourteen Western battery companies collapsed between January 2025 and April 2026, despite raising over $20 billion between them. The post-mortems — bad management, premature scaling, single-customer risk — are accurate but incomplete. The deeper cause was a shared belief that capital could substitute for time in a physical process industry. It cannot, and this edition explains why.
You will also understand why China’s state-led industrial policy, which drove lithium prices down 70 to 80% from their 2022 peak, made every Western execution mistake lethal; why targeting the automotive market was the worst possible entry point for companies still learning to manufacture; and why the gap between Western startups and Chinese incumbents was never primarily a capital problem.
The more important argument, however, is what the graveyard leaves behind. A nucleus of hundreds of engineers across Europe holds process knowledge that did not exist a decade ago. This edition sets out four conditions that could turn that foundation into something, the financing structures emerging to support it, and the historical precedents — from Taiwan’s semiconductor industry to postwar European aviation — that show what coordinated action from such a starting point can achieve.
What went wrong, and why
The Battery Chronicle recently catalogued the wreckage of fourteen Western battery companies and identified five recurring patterns across the failures: companies that scaled before their first plant worked, companies that never reached real revenue, poor management and sloppy execution, total concentration risk on a single customer, and value chains that stopped halfway.
Each pattern is real, each post-mortem is worth reading, and taken together they form a coherent account of what went wrong at the operational level. We believe, however, that a large part of the explanation sits one level up, and that the wreckage left something behind that the operational account tends to miss.
The five patterns share a common root. Most of these companies attempted to reach China-style gigascale production before proving the minimum viable unit of industrial production, which in battery manufacturing is a single working small GWh line — one line running stably and delivering product. In retrospect, the pressure to reach that point while simultaneously managing customer relationships, building a supplier base, constructing infrastructure, and satisfying demanding investors created conditions in which good management was nearly impossible, and bad management was almost inevitable.
Having opted for the automotive market added a further layer of difficulty. True, there was no obvious alternative at the time: automotive was the only market large enough to justify allocating venture capital at the required scale, which meant that battery startups had little choice about where to aim. But automotive batteries are also the most demanding first product imaginable. They require cycle-life above 1,000 charges, an eight-year warranty, and large-format cells produced at high volume with near-zero defect rates. In the first years of a new production line, scrap rates typically run at 15 to 30 percent — and can reach 70 to 80 percent in the worst cases — with each percentage point costing roughly €10 million per year at a line running at full capacity, according to research by Fraunhofer FFB and RWTH Aachen University.
By contrast, LG and Samsung, whose trajectories are often cited as benchmarks, spent the better part of a decade on batteries for camcorders and laptops before they attempted anything like this. Compared to batteries made for EVs, it made a big difference: batteries for smaller devices require 300 to 500 charge cycles, a two-year warranty, and small cells, with development cycles of roughly one year that allow for rapid iteration and learning. The knowledge Asian cell manufacturers accumulated over a decade of lower-stakes production is precisely what Western startups tried to acquire in five years — on the hardest possible product, with investor timelines that left no room for the kind of grinding, iterative process work that actually builds manufacturing capability.
Managing the supply chain was an additional challenge. From what we’ve learned, Chinese equipment suppliers deliver roughly 70% of the machinery needed for a battery production line, but the remaining 30% is proprietary process knowledge held internally by companies like CATL and LG — knowledge that does not transfer with the purchase order and cannot be acquired from any equipment catalogue. There are multiple examples of European companies who bought Chinese equipment and required more than a year and dozens of engineers embedded at suppliers to redesign it before it functioned as intended. The machine, in other words, is only part of the answer: the knowledge of how to run it, adapt it, and integrate it into a functioning production system lives in people, and it accumulates through multiple iterations that cannot be skipped or purchased.
China, and the trap of investing into a boom
Capital cycle theory explains what happened in the battery industry over the past decade: when everyone invests at once, excess supply follows and returns suffer. The more specific warning embedded in the framework is to track competition at the level where it actually occurs, because investors who focus on local or regional dynamics while a global competitor is reshaping the economics of the entire sector will consistently misread the cycle. How have those investors missed China and the formidable challenge it represented for their future returns?
As in most markets, the scale of what China did in batteries, EVs, solar, and semiconductors over the past decade is simply unprecedented. Roughly $1.4T of state capital was deployed into these sectors, with a goal that — as always with China — was not short-term financial return but long-term industrial dominance. Beijing’s typical playbook is to flood targeted sectors with capital, accept colossal waste as a feature rather than a bug, and allow brutal domestic competition — what Chinese officials now call neijuan, or ‘involution’ — to sort out a small number of globally dominant survivors from a very large field of entrants. In this case, roughly 500 companies entered the Chinese EV market; around 150 to 200 remain today, most still loss-making; and consolidation has further to run, with perhaps ten eventually surviving the full cycle. The recently established National M&A Guidance Fund is now tasked with managing that process, absorbing failing firms and recycling their talent and intellectual property into the national champions that Beijing has decided to back.
Because China’s approach to industrial policy simultaneously suppresses domestic consumption — keeping interest rates artificially low and directing lending toward industry rather than households — Chinese manufacturers cannot absorb their own output domestically, and the surplus is exported. As a result, China ran a record $1.2T trade surplus in 2025, with EVs and batteries contributing significantly to that figure. The downstream consequence for Western battery companies was direct and severe: lithium prices fell 70 to 80% from their 2022 highs. That collapse gutted the recycling business model at the moment these companies most needed it to hold. And while lithium prices have since recovered, Europe failed in the interim to build the hydrometallurgical recycling capacity its own battery regulations now require, and black mass is leaking out to China to fill the gap.
The three recyclers in the graveyard — Li-Cycle, Lithion Technologies, and Ascend Elements — had all promised investors the same full-circle story: collect old batteries, break them into black mass, and refine that material into battery-grade lithium, nickel, and cobalt. None of them finished the refining step, for one simple reason: lithium recovery from black mass is technically very difficult, and yields are hard to optimise for a material whose price is volatile and effectively controlled by China. Indeed, the economics of the full-circle model depended on lithium prices that no longer existed by the time the plants were supposed to be running.
Against this backdrop, the accumulated advantage of CATL and BYD explains where we are today. Both companies spent fifteen years building operational knowledge one production cycle at a time before committing to multi-gigascale production, and BYD’s vertical integration across batteries, motors, and power electronics represents process knowledge compounded across that entire period. That kind of advantage cannot be acquired through a funding round or replicated through capital deployment, because it is built into the people, the processes, and the production systems that took fifteen years to develop. Western companies that entered the market at exactly the moment when this advantage was most fully realised, and when China’s export-driven overcapacity was pushing input prices to historic lows, were operating in conditions that made every execution mistake lethal.
You cannot buy a learning curve
The capital involved in the Western battery push was modest by industrial standards — mostly venture-scale, with the notable exception of Northvolt, which raised $15B before its collapse. This was not a financial bubble in the classical sense: the total capital deployed was relatively small, and the mechanisms that typically inflate and then puncture financial bubbles were not the primary driver of what happened. Instead, it now appears Western industrials succumbed to arrogance.
The underlying belief was that the gap between Western startups and Chinese incumbents could be closed with sufficient capital and sufficient speed — that what China had achieved through years of heads-down, grinding process work was essentially a compression problem, and that money could compress it. Part of what made this belief plausible was a misjudgement about the nature of the competition: China was seen as a beatable adversary if the right resources were deployed fast enough, rather than as a manufacturing civilisation that had been building process knowledge in this specific domain for a decade and a half.
In the European context, the green transition argument added a further dimension, giving the entire effort a sense of moral purpose that may well have reinforced the confidence that this time, with these motivations, the outcome would be different. The approach — repeat what China did, but faster and with better environmental credentials — turned out to be mistaken. Physical process industries do not respond to capital the way software businesses do. The learning curve is paid in time and iterations, and there is no mechanism by which funding rounds can substitute for either.
Luckily for the West — and Europe specifically — the investment cycle that just ended was not purely destructive. Coesia, the Italian equipment maker, developed battery assembly technology through its partnerships and has since won orders from Agratas for its planned gigafactory in India — a European equipment champion with genuine global reach that would not exist without the capital that flowed into the sector over the past decade. Likewise, Dürr and other suppliers in the Stuttgart region developed electrode process technology, including dry coating techniques, that had been considered impossible to achieve in Europe until Porsche’s Cellforce demonstrated otherwise, using suppliers drawn entirely from the Stuttgart region; other OEMs in Bavaria have since recognised that capability as real and replicable. Enough engineers and technicians have now been through a full production cycle to constitute the foundation of something new, if the conditions are right.
Europe can find further solace in an interesting Taiwanese precedent. In 1973, Taiwan had no semiconductor industry to speak of. The government established the Industrial Technology Research Institute (ITRI) and later sent a small team of engineers to RCA’s facilities in the US to acquire hands-on knowledge of chip design, manufacturing processes, quality control, and equipment management. They returned, built a demonstration factory that outperformed RCA’s own plants, and went on to spin out TSMC and UMC, transferring the key engineers and the demonstration facility with them. The entire Taiwanese semiconductor industry, which today produces the most advanced chips in the world at a massive scale, traces back to that nucleus of a few hundred people who held integrated, end-to-end process knowledge and applied it in a concentrated, deliberate way.
That said, time is of the essence, and the accumulated process knowledge won’t stay here forever. When France stopped building nuclear reactors in the 1980s, experienced workers retired without successors, the rhythm of continuous production broke, and the knowledge that had made the French programme one of the most efficient in the world dispersed into retirement and other industries. Modern French companies now struggle to execute new reactor builds because they are selling technology without the industrial ecosystem that once made it efficient. Boeing provides a more recent version of the same lesson: the decision to strip out veteran engineers and quality inspectors in the name of cost reduction permanently destroyed the process knowledge that had ensured its products worked, with consequences that are still unfolding.
When it comes to batteries, a nucleus of 200 to 300 people with integrated process knowledge — from lab scale through to industrial production — exists today in Europe, concentrated in the Stuttgart region with some in Munich and northern Italy. That nucleus is precisely what the Battery Chronicle‘s graveyard analysis does not account for. In our case, the graveyard is also a foundation, and the question of whether anyone builds on it deliberately, before the knowledge walks out the door, is more consequential than any of the individual post-mortems.
The conditions for a restart
Four things need to align for that foundation to become something.
1️⃣ The first is cycle speed.
Here, the lesson from LG and Samsung’s trajectory is directly applicable. The right initial markets for a European battery restart are those with smaller volumes and lower capital requirements — defense, data centers, and other performance-driven applications — where it is possible to deliver the product sooner without needing Chinese-style gigascale to be cost competitive. Smaller volumes mean lower capital risk and faster proof of concept, and faster proof of concept means faster accumulation of the process knowledge that is the actual source of competitive advantage.
These markets also create demand for performance specifications that differ from and in some cases exceed automotive requirements, and supply chain sovereignty has become an explicit procurement criterion in both sectors rather than a political aspiration. The EU’s Carbon Border Adjustment Mechanism (CBAM) adds a structural cost advantage for European producers running on clean energy — Nordic hydro or French nuclear — over carbon-heavy imports. The path to automotive competitiveness runs through these markets, as costs fall and the production ecosystem compounds, rather than through automotive from the start.
2️⃣ The second condition is demand coordination.
Single-customer dependency is a solvable problem, but only through coordination that did not happen in the previous cycle. CustomCells was locked into an Airbus-style supplier contract with Lilium that assumed production-ready technology while leaving CustomCells to bear the development costs; Cellforce was constrained throughout its existence by the same dynamic. By contrast, the right model aggregates demand across sectors, with defense procurement bodies, hyperscalers, and automotive OEMs acting as joint rather than competing sources of committed demand.
The precedent that best illustrates what this looks like in practice is Airbus itself: in the 1960s, France, the UK, and Germany pooled resources to bear the full development cost of a European aircraft, secured early subsidised market commitments that allowed the consortium to survive its early years, and eventually locked in the major orders that made the programme self-sustaining.
Today, the Electric Tech Stack — batteries, motors, power electronics, and semiconductors — has become the foundation of modern warfare in a way that makes defense procurement a particularly compelling anchor: drones accounted for 60 to 70% of all equipment losses in Ukraine, and they run entirely on this stack. China’s recent implementation of export controls on batteries above 300 Wh/kg has transformed local sourcing from a preference into a strategic necessity. On the data center side, 68% of new hyperscale projects incorporated on-site or microgrid power in 2025, up from 29% in 2021, driven by grid connection queues that now stretch up to five years and by the reality that even millisecond power interruptions can corrupt weeks of computation.
3️⃣ The third condition is the right financing structure.
Venture capital is too impatient and too small in ticket size for gigafactory build-out; meanwhile, infrastructure private equity finds the risk profile too early-stage for its mandates; finally, corporate investors want ownership and control that most founders cannot accept. European policy acknowledges the market exists but cannot fill the gap through grants alone, and America’s Inflation Reduction Act demonstrated in the US that even billions of public capital are insufficient if the private financing structure is wrong.
The solution that is emerging in adjacent sectors is what is now known as production capital: venture capital for the early stage of go-to-market, government-backed debt for factory build-out, with grants, pre-orders, strategic partnerships, and project finance orchestrated together rather than sequenced one after another. The Apollo/8VC model — patient asset-backed financing combined with startup operational expertise — points in the right direction. Forge Nano’s $100 million Department of Energy award to build a 3 GWh defence-focused battery facility in the US illustrates the template well: secure high-value sovereign demand first, use it to fund the learning curve, and expand from there into broader commercial markets once the production system is proven.
4️⃣ The fourth condition is the protection and maintenance of the core.
The 200 to 300 people who hold integrated process knowledge need to be identified, funded, and kept together around a minimum viable 1 GWh line — the smallest unit of real industrial scale, and the base from which 10 to 20 GWh and a high-value niche segment would eventually make automotive cost competitiveness achievable.
The historical analogy that captures what is at stake is the small group known as the “Traitorous Eight”: in the early 1960s, eight engineers left Shockley Semiconductor, founded rival company Fairchild, and later Intel, carrying with them the foundational process knowledge that seeded the entire American semiconductor industry. What this suggests is that the core team is absolutely critical in seeding a new industry, as it acts as the demand function for the entire supply chain that surrounds it — the electrode makers, the assembly equipment specialists, the process technology firms in Stuttgart and northern Italy. Without such a core team, that supply chain has no anchor customer, no source of iterative feedback, and no path to the future.
Allied scale
Europe has the mass. Process knowledge, supply chain depth, trained people, production experience, and a structural cost advantage under CBAM — these exist today as a direct inheritance from the investment cycle that just ended, and most of those would not exist without it. The question that remains is whether the mass can be converted into coordinated scale, which is a different problem entirely.
American capital is already moving into European industrial assets, deal by deal and without a coordinating logic — while the EU's own tools for managing Chinese investment in the same assets are proving slow and divisive. KKR is acquiring German industrial companies. Carlyle agreed to buy BASF’s coatings division for €7.7 billion. IonQ acquired Oxford Ionics for approximately $1.1 billion. Atomico led a €34.5 million Series A in Lace Lithography, a Norwegian-Spanish company developing atom-beam lithography systems for chip production, with participation from Microsoft M12. (Lace Lithography is part of the Vsquared portfolio.) The flow of transatlantic capital toward European productive assets is real and growing, driven by the recognition that European industrial heritage — the process knowledge, the engineering depth, the supply chain integration — represents something that American capital can work with in ways that purely domestic deployment cannot replicate.
Rush Doshi’s framework of “allied scale”, developed in Foreign Affairs article with Kurt Campbell, provides the right vocabulary for what this moment requires. The Western alliance as a whole holds roughly three times China’s nominal GDP and about twice its GDP on a purchasing power parity basis, but mass of that kind only translates into leverage if it is organised and coordinated rather than left to accumulate in disconnected national silos.
The postwar period offers a precedent: Jean Monnet and the American “Wise Men” who worked alongside him understood that European mass had to be converted into coordinated scale through deliberate institution-building, and the result was an alliance that held for decades. The current situation has some of the same ingredients — American capital, European productive capability, shared strategic interests in batteries, defense, and data center infrastructure — but lacks the coordinating institution or initiative that would allow them to function as a system rather than as a collection of individual transactions.
A restart in European batteries is one concrete place to begin building that coordination, and it has the advantage of being tractable: the people exist, the supply chain exists, the demand signals are visible, and the financing models are being developed in adjacent sectors. What it requires is the kind of deliberate alignment — among demand, capital, and the teams that hold the knowledge — that has historically required either a crisis or a visionary institution to achieve.
Conclusion
The graveyard list documented by the Battery Chronicle is an excellent starting point for this endeavor. Assets are already changing hands: Glencore operates Li-Cycle’s collection facilities, Lyten acquired parts of Northvolt, SKion rescued BMZ, and private investors acquired CustomCells, which now produces battery cells for military applications. The technology in most of these companies worked. What the post-mortems do not ask is what the failures left behind, and the answer to that question is more consequential than any individual accounting of mismanagement or concentration risk. Process knowledge, supply chain capability, trained engineers, and equipment makers with global reach constitute a foundation that did not exist ten years ago. The next cycle, if it is approached with more realism about what physical process industries require, can build upon it.
In other words, the learning curve cannot be bought, but it can be inherited. And in the Stuttgart region and northern Italy, the inheritance is there. Existing sites, technology, and people could be consolidated into a new entity at a fraction of the cost of the original investments — a kind of EU Airbus for batteries, built not from scratch but on the foundations that the previous cycle laid down. What that requires is not another wave of venture capital optimism but the kind of deliberate, coordinated industrial policy that turned European aviation from a collection of national also-rans into a genuine global competitor. Whether the demand, the capital, and the institutional will can be aligned before the window closes is the question the graveyard, read correctly, leaves open.
Cylib received a €63.4 million funding certificate from the German federal government and North Rhine-Westphalia for its battery recycling facility in Dormagen, presented at the State Chancellery in Düsseldorf by Minister-President Hendrik Wüst and Parliamentary State Secretary Gitta Connemann. LINK
IQM will deploy a 20-qubit quantum system to TOYO Corporation by end of 2026, marking the first enterprise-owned quantum computer in Japan and extending IQM’s presence in Asia-Pacific alongside existing deployments in South Korea and Taiwan. LINK
Make Europe the Electro Union (Norrsken). An open letter coordinated by Norrsken and signed by a coalition of European investors and corporates calling for the EU to commit to running over 50% of its economy on clean, domestic electricity by 2040. The argument: three fossil fuel price shocks in four years, costing an estimated €930 billion in crisis premiums since 2021, have made the case for electrification unanswerable. The letter frames cheap, abundant domestic electricity not as an energy policy but as the foundation on which every European industrial company of the next twenty years will be built. LINK
The World’s Most Complex Machine (Works in Progress). A long read on how ASML built its EUV lithography monopoly — the thirty-year learning curve, the supplier ecosystem, and the co-investment model with customers that made it possible. Almost a direct analogue to the battery manufacturing story told in this edition, and direct context for the Lace Lithography portfolio news above. LINK
1 to 100 in China: Inside Beijing’s Scaling Machine (Cam Watson). A close-up account of how local government officials, industrial clusters, and state incentives carry companies from first product to scale in China — with a speed and co-ordination that European founders consistently describe as having no equivalent here. The operational complement to the strategic argument this edition makes about what coordinated industrial policy actually looks like on the ground. LINK
How Ukraine Scaled to Millions of Drones (ChinaTalk). Jordan Schneider and Chris Miller interview Cat Buchatskiy of Snake Island on Ukraine’s drone industrial base: from 3,000 units at the start of the war to millions per year, built largely on Chinese components. The piece maps supply chain dependencies, the role of battlefield feedback loops, and what rapid iteration in a physical process industry actually requires — all directly relevant to the Electric Tech Stack argument in this edition. LINK
No, America Is Not in a Stealth Manufacturing Boom (Noahpinion). Noah Smith pushes back on a Wall Street Journal piece arguing that US manufacturing is quietly recovering, making the case that tariffs are cancelling out the tailwinds from the IRA and CHIPS Act. Worth reading alongside this edition’s argument about what coordinated demand and capital can and cannot achieve — and as a caution against assuming that policy intent translates into industrial outcome. LINK
Why Europe, Why Now (Techno-futurism). David Ordonez of the NATO Innovation Fund argues that geopolitical pressure, maturing deep tech venture capital, and a new generation of science-rooted founders are creating the conditions for a European industrial renaissance — grounded in research, precision manufacturing, and engineering depth. A constructive counterpoint to the graveyard narrative, and the optimistic frame this edition ends on. LINK
