Weekly: Europe's semiconductor strategy
15.5 min read.
Highlights
Europe’s semiconductor strategy. The Nexperia debacle both coincides and fuels a lot of commentary today about Europe’s semiconductor industry. Chris Miller, author of Chip War, and General John R. Allen of the US Marine Corps (Ret.) together write a piece in Foreign Affairs about Europe’s sliding position in the semiconductor race. Europe is not without its stars, ASML and Arm to give two examples, has plenty of room for collaboration with the United States in auto and defence industries, beyond the headline-grabbing AI chip space.
The FT Editorial Board and the Economist both react to the Nexperia situation to call for a better European chip strategy.
FT on the AI bubble. Following last week’s commentaries on the financial network of AI/semiconductor deals and the ongoing wider debate around the AI boom/bubble, there are several articles in the FT discussing 1) OpenAI’s US$1 trillion worth of deals for a chipmaker make a fraction of that in revenue, 2) the increased financialisation of the AI boom, and 3) the difference in pricing between buying GPUs ($50,000/GPU) and renting one ($2.80/hour).
WSJ on the AI bubble. The WSJ also has a few articles on the AI bubble/boom debate, delving into 1) the Broadcom / OpenAI deal and 2) watching the demand for AI.
Thanks for reading.
Weekly Chip Briefing: October 11 – October 17, 2025
Chris Miller and John Allen, “Europe Is Losing the Chips Race,” Foreign Affairs, 10/16/2025.
FT Editorial Board, “Europe needs a better chip strategy,” FT, 10/15/2025.
The Economist, “The Dutch seize control of Nexperia from its Chinese owner,” The Economist, 10/16/2025.
Richard Waters, “How OpenAI put itself at the centre of a $1tn network of deals,” FT, 10/11/2025.
FT Editorial Board, “Measuring risk in the AI financing boom,” FT, 10/14/2025.
Bryce Elder, “What GPU pricing can tell us about how the AI bubble will pop,” FT, 10/17/2025.
Asa Fitch “Why Broadcom’s Bet on OpenAI Is a Big Risk,” WSJ, 10/14/2025.
Steven Rosenbush, “AI Economics Are Brutal. Demand Is the Variable to Watch.,” WSJ, 10/14/2025.
Winston Ma, “Technological sovereignty with American characteristics,” FT, 10/12/2025.
1.
Chris Miller and John Allen, “Europe Is Losing the Chips Race,” Foreign Affairs, 10/16/2025.
European leaders have grand ambitions to reduce the continent’s reliance on sensitive technologies from abroad. Today, they are debating an update to the European Chips Act, which was finalized in 2023 and allocated billions of euros to subsidize chip-making on the continent. The act was meant to increase Europe’s share of global chip manufacturing from ten to 20 percent by 2030, but it will likely fall short of that target by a wide margin.
A purely European supply chain for semiconductors—the sector that undergirds the digital economy and defense sector—is a fantasy that distracts from real opportunities. Many of the powerhouses of Europe’s chip industry, such as ASML, a Dutch company that makes semiconductor equipment, and Merck, a German firm that produces chemicals for chip-making, don’t manufacture semiconductors. Companies like these are cutting-edge, are often highly profitable, and draw on Europe’s industrial expertise in precision machinery, specialty chemicals, and advanced materials. Yet they are overlooked by politicians who focus on chip output.
At the same time, the semiconductor industry is facing rapid technological changes, trade restrictions, and geopolitical shifts that Europe is not prepared for. The main driver of rising demand for chips is artificial intelligence, an area in which Europe is comparatively weak. European firms are being squeezed out of China’s market, which used to be a primary source of growth for them. And the United States, which used to be a close partner, is threatening tariffs that would limit European sales to the U.S. semiconductor market. Merely setting targets for domestic chip production will not solve these challenges.
A self-sufficient semiconductor industry may be out of Europe’s reach, but a more vibrant one is not. Europe has a meaningful edge in certain steps of the semiconductor supply chain, and it can cooperate with allies—including the United States—for the supply chain segments it lacks. The billions of euros being poured into the continent’s rearmament can also be an opportunity for its chip makers, given how critical artificial intelligence has become to defense. To take full advantage of these trends, European leaders need to build on the chip industry’s strengths with deeper partnerships, not a futile drive for self-sufficiency.
2.
FT Editorial Board, “Europe needs a better chip strategy,” FT, 10/15/2025.
The Dutch government’s seizure of chipmaker Nexperia from its Chinese owner Wingtech is a landmark moment in Europe’s evolution from one of the world’s most open trading blocs to one increasingly preoccupied by its economic security.
It is all the more striking coming from a small free-trading country with laissez-faire business instincts which approved the sale of Nexperia to Chinese investors in 2017 — a decision that even with the benefit of hindsight it must regret. With this takeover, the Netherlands has stepped straight into the struggle for technological supremacy between the US and China centred on the semiconductor industry.
Nexperia was one of many large-scale strategic takeovers by Chinese investors of western companies in critical technologies or infrastructure during the middle of the last decade. It is the first to be fully clawed back by a western government. It may not be the last.
European countries have intervened with Chinese owners before. France seized the shares of Chinese-controlled chipmaker Ommic in 2023. The previous year the UK government ordered Nexperia to sell the bulk of its stake in Newport Wafer Fab. Both Ommic and NWF were small outfits producing or potentially producing more sophisticated semiconductors for military applications.
The Netherlands made a mistake allowing Nexperia’s sale, even if it has now moved decisively to correct it. The episode underscores the need for the EU to develop a better semiconductor strategy. It is way off its ambition for a 20 per cent global market share. Thinking through its dependencies would be a good start.
3.
The Economist, “The Dutch seize control of Nexperia from its Chinese owner,” The Economist, 10/16/2025.
AT THE HEIGHT of the cold war, the Netherlands passed a law allowing the state to take over companies for security purposes. After gathering dust for 73 years, the law has at last been put to use. On September 30th the Ministry of Economic Affairs quietly took control of Nexperia, a semiconductor firm headquartered in Nijmegen that had been bought in 2019 by Zhang Xuezheng, a Chinese entrepreneur. A week later an Amsterdam business court suspended Mr Zhang (known by his nickname “Wing”) as CEO, replacing him with a Dutch interim chief. The moves were made public by the Dutch press on October 12th. They are among the most aggressive steps yet by European governments to protect strategic industries from China.
The Dutch were prompted in part by fears that Mr Zhang was undermining Nexperia in favour of his Chinese companies, and in part by threats from America. Last year the government told the firm it was at risk because America’s Department of Commerce had put Wingtech, Mr Zhang’s holding group, on a sanctions list. On September 29th the Americans extended export controls to companies whose majority owners are on the list. Mr Zhang had reportedly frustrated negotiations on guaranteeing Nexperia’s independence.
Meanwhile, changes at Nexperia were setting Dutch executives’ hair on fire. The company had become a regular client of a chip foundry in Shanghai that Wingtech set up in 2020. That foundry had run into financial problems, the Amsterdam court said, citing claims by Dutch executives that Mr Zhang was trying to divert cash from Nexperia to prop it up. In September the firm’s chief financial officer and two other executives found their access to company accounts had been revoked and given to people with little financial experience. Another senior financial officer quit, writing in an email that Nexperia had become “purely Chinese”.
4.
Richard Waters, “How OpenAI put itself at the centre of a $1tn network of deals,” FT, 10/11/2025.
The pursuit of a humanlike artificial intelligence does not come cheap. OpenAI, the company that sparked the tech industry’s AI race, recognises that its quest will need more capital than any investment project in corporate history. As a result, it has been forging new and unusual financial bonds with some of the biggest names in tech.
The size of that gamble has become clear in recent weeks as the company behind ChatGPT has lined up a series of deals that could lead to it spending more than $1tn on computing power.
Tapping into that much capital has led OpenAI to weave deals that draw on the financial resources of other Big Tech companies, adding to a growing web of financial dependencies across the AI world. In the process, it could be helping to create a new level of systemic risk in an industry that may already have entered bubble territory.
Parsing the risks from giant transactions like these has brought a new complexity to the financing of the AI boom.
One challenge has been weighing the odds that these and other megadeals will ever be fully consummated. Among the unknowns: whether demand for AI services will be strong enough to justify building all the data centres, whether the new facilities can be financed, built and equipped, and whether there will be enough electricity to power them.
Beyond the sheer scale and uncertainty of contracts like this, meanwhile, there are also questions about whether they are leading to new interdependencies and wider systemic risks in the AI world. The circularity of the way money moves between companies in some transactions, for instance, could inflate the apparent level of demand.
With more interdependence, a setback at one big AI player could reverberate through the industry. And, depending on how they are financed, the effects could spread beyond the AI world, putting a dent in the wider financial system.
5.
FT Editorial Board, “Measuring risk in the AI financing boom,” FT, 10/14/2025.
The artificial intelligence boom is entering a new and riskier phase. Global capital expenditure on chips, data centres and cloud infrastructure has, so far, been driven by hyperscalers, funded largely through their vast internal cash balances. But the projected scale of computing power needed for generative AI is now prompting a shift towards more leveraged, opaque and circular financing structures, which raises the economic stakes riding on the technology’s success.
Between now and 2029, global spending on data centres is forecast to hit almost $3tn, according to Morgan Stanley. Big Tech groups will remain major spenders, but cash flows from their advertising and cloud computing businesses can only stretch so far. Many are increasingly looking to public and private credit sources to support further expansion. In August, Meta raised $29bn — including $26bn of debt — from private capital investors to fund centres in Ohio and Louisiana.
Even if the AI build-out proves to be overzealous, history suggests the excess could leave behind useful infrastructure, much like the railway boom of the 19th century or the fibre networks laid during the dotcom bubble. A glut of cheap compute power could then support the next wave of growth and innovation.
Either way, the data centre investment surge is no longer just a bet on productivity, but increasingly a test for financial stability as well. With credit, capital and confidence set to be more entwined, a stumble could shake far more than tech valuations. Corporate discipline, investor scrutiny, and regulatory vigilance will now be more important in ensuring the AI boom builds more than it breaks.
6.
Bryce Elder, “What GPU pricing can tell us about how the AI bubble will pop,” FT, 10/17/2025.
One odd thing about AI equipment is that it’s very expensive to buy and very cheap to rent.
Want an Nvidia B200 GPU accelerator? Buying one on its release in late 2024 would’ve probably cost around $50,000, which is before all the costs associated with plugging it in and switching it on. Yet by early 2025, the same hardware could be rented for around $3.20 an hour. By last month, the B200’s floor price had fallen to $2.80 per hour.
Nvidia upgrades its chip architecture every other year, so there’s an opportunity for the best-funded data centre operators to lock in customers with knockdown prices on anything that’s not cutting edge. From the outside, the steady decline in GPU rental rates resembles the kind of predatory price war the tech industry relies upon: burn money until all your competitors are dead.
Price wars in the real world often involve small companies undercutting big ones, but it’s an upside-down version of how tech usually works. Why? We don’t know, but here’s a guess.
GPU-as-service customers have historically been AI start-ups and research institutions wanting to train new models, meaning they need lots of computing power for a relatively short period. They may already be customers of the hyperscalers, so staying with the same host may have continuity, efficiency and security benefits that justify the premium.
The second type of customer is the regular corporate that wants a website chatbot, summarisation tools or similar AI widgets. Only very big and/or paranoid organisations will want to manage the required infrastructure, so everyone else might in the past have used a GPU in the cloud. Now, however, they’re much more likely to build chatbots etc using a ready-made LLM like OpenAI or Anthropic and pay by the token rather than by the hour.
7.
Asa Fitch “Why Broadcom’s Bet on OpenAI Is a Big Risk,” WSJ, 10/14/2025.
Broadcom’s agreement to develop huge numbers of chips and computing systems with OpenAI is being welcomed by investors. But it isn’t such a clear-cut win as the company’s stock is suggesting.
The deal capped a flurry of pacts between OpenAI and the world’s biggest AI chip suppliers, including Nvidia and Advanced Micro Devices. All of them involve OpenAI spending billions of dollars on data centers stocked with hundreds of thousands of chips. Broadcom’s stock rose 9.9% Monday following news of its deal.
The thing is, it isn’t clear how OpenAI is going to pay for all of it, including the deal with Broadcom. OpenAI’s revenue this year is expected to be around $13 billion, a substantial sum for a startup but nowhere close to enough to justify Altman’s exuberance. The company has told investors it won’t be profitable until 2029.
For Broadcom, though, it amounts to a big bet on a chancy customer. It could pay off handsomely: Bernstein Research analyst Stacy Rasgon estimated Monday that well in excess of $100 billion of additional revenue could be possible in the next three to four years from the deal. But it is a gamble, and there is reason to doubt whether Altman will achieve anything near the full extent of his ambitions.
In this sense, Broadcom is making an even bigger bet on OpenAI than Nvidia or AMD. Those two are still competing for its attention, though, which could fuel price competition and further margin pressure for Broadcom. In a sign that Broadcom’s custom-chip business already is under competitive threat, Alphabet’s Google, the business’s anchor customer, has started working with Taiwan’s MediaTek on custom AI chips.
Investors seem content to ignore such concerns, as they have been for some time. Broadcom is essentially a mishmash of chip and software businesses that don’t often intersect, yet investors have been willing to pay a premium for its shares: They now trade at about 40 times next year’s earnings.
Cozying up with OpenAI could set the stage for years of booming sales at Broadcom. With OpenAI’s future riding on a grandiose vision that lacks a clear financial model, though, investors shouldn’t see it as anything close to a sure thing.
8.
Steven Rosenbush, “AI Economics Are Brutal. Demand Is the Variable to Watch.,” WSJ, 10/14/2025.
AI companies are losing money at an epic pace, and the reasons go deeper than mere profligacy. The economics of artificial intelligence have turned sharply against them, at least for now, and for reasons that weren’t widely anticipated.
It’s hard to predict how this will play out in the financial markets, but here’s a clue. Keep an eye on demand for AI, measured in units of data processed. It’s soaring right now. The entire AI bet may turn on how far and fast it ramps from here.
A key mistake occurred a year ago when AI leaders focused solely on AI cost curves for a unit of computing and took their eye off the number of units needed to get the work done, said Heath Terry, the global sector lead for technology and communications research at financial services firm Citi. AI computing costs had been declining around 90% every seven months, a dynamic akin to Moore’s Law for microchips, giving AI companies reason to believe their price-performance ratio will improve.
Then usage, which everybody expected to increase as the price fell, went absolutely parabolic.
The demand for tokens has been driven in part by the fight against the “hallucinations,” or incorrect or nonsensical outputs stated as if they are true, notoriously produced by generative AI models. Current cutting-edge models are designed to counter the problem often with the help of a kind of reinforcement learning in which they answer the same question multiple times internally before issuing a response. They also might query a so-called mixture of experts, meaning specialized models. That has helped AI produce better answers—and exacerbated AI’s resource needs.
So the unit price of a token has fallen, yes, but the overall consumption of tokens has gone nuts.
That demand will only continue to soar. OpenAI’s new Sora short-form AI video app powered by the latest version of its Sora video generator is a hit. (News Corp has a content-licensing partnership with OpenAI.) The company is looking to ramp up efforts in AI commerce as well. And Anthropic’s new large language model, Claude Sonnet 4.5, can code for 30 hours straight. Such longer-thinking models are an emerging form of AI that will put the industry’s ballooning data-center capacity to use.
9.
Winston Ma, “Technological sovereignty with American characteristics,” FT, 10/12/2025.
The Trump administration’s equity-for-grants deal with Intel did three things, each of which sheds light on America’s evolving industrial strategy.
First, it sent a market signal that the US is committed to Intel’s long-term prospects. The company is the country’s best chance to compete with Taiwan Semiconductor Manufacturing Company and Korea’s Samsung in chip fabrication.
Second, the stake was a “poison pill” to dissuade the company from fully exiting the manufacturing sector. TSMC is the dominant player in chip manufacturing, especially advanced artificial intelligence chips, so Wall Street may advise Intel to focus on chip design instead. But Intel pulling out of manufacturing would be detrimental to the government’s efforts to shore up domestic chip making for reasons of supply-chain stability.
Third, the Intel deal was clearly intended to “crowd in” further public-private partnership. Within days of the US government taking its stake, Japan’s SoftBank announced its own $2bn investment in Intel, followed by Nvidia’s $5bn design and manufacturing partnership with the company.
Will the US approach succeed? It’s worth comparing these recent ad hoc efforts to boost chip manufacturing in America with China’s National Integrated Circuit Industry Investment Fund, which demonstrates Beijing’s ambition to build a competitive domestic foundry ecosystem. After significant investment by NICIIF in Shanghai-based Semiconductor Manufacturing International Corporation, the company now makes advanced AI chips for Huawei, and is striving to catch up with TSMC as well.
Both the US and China have decided that they can’t afford to outsource their chipmaking future. Technological sovereignty is their common destination, and they are resorting to the same kind of strategy, albeit with their own distinctive characteristics, to get there.
-