Last Updated on August 17, 2026 by Karl Thompson
In late October 2025, Honda stopped production at an assembly plant in Mexico. Across the Atlantic, German automotive suppliers began preparing applications for short-time working. Orders were not the problem. Orders were fine.
The problem was a component costing a few pence, made by a company most drivers have never heard of.
Nexperia is a Dutch chipmaker, headquartered in Nijmegen and owned since 2019 by the Chinese firm Wingtech. It produces around 110 billion chips a year — unglamorous things, transistors and diodes, the kind used to switch lights on and off in a car — which amounts to roughly a tenth of global supply in certain product categories and around forty per cent of what the European automotive industry needs.
The sequence that stopped the Honda line did not begin in Nijmegen or in Mexico. It began in Washington. On 29 September 2025, the US Bureau of Industry and Security extended its Entity List, with immediate effect, to cover any foreign affiliate controlled by a listed company — which captured Nexperia, because Wingtech had been listed in December 2024. One day later, the Dutch economic affairs minister invoked a Cold War-era statute to take temporary control of the company. Four days after that, China’s Ministry of Commerce halted exports of the chips finished at Nexperia’s plant in Dongguan.
A rule change in Washington, a ministerial order in The Hague, a customs decision in Beijing, and a stopped line in central Mexico, inside four weeks.
This article is part of How Modern Society Works, a series on the systems that organise everyday life. Each article takes something ordinary and works outwards to the system that produced it.
How Is a Semiconductor Actually Made?
Three stages, and the distinction between them is the whole subject.
Design is where a chip’s architecture is worked out and validated. It needs specialised software, accumulated intellectual property and expensive engineers, and almost no factory.
Wafer fabrication — the front end — is where the design is physically etched onto silicon discs, hundreds or thousands of copies at a time, through hundreds of precisely controlled steps. This is the part requiring a multi-billion-pound plant, ultra-pure water and a cleanroom.
Assembly, testing and packaging — the back end, usually abbreviated to ATP — is where the wafer is cut into individual chips, each working one is encased in a protective housing and bonded to its connections, and the finished device is put through electrical and burn-in tests to catch failures before a customer does.
Firms sort themselves along those stages. Integrated device manufacturers do all of it. Fabless design companies do only the first and buy the rest. Foundries do only fabrication, for other people’s designs. Outsourced assembly and test houses do only the back end.
Nexperia is close to an integrated manufacturer, but split across continents: wafers fabricated in Europe, most packaging and testing done in China. When the two halves stopped shipping to each other, the front end had wafers it could not finish and the back end had capacity it could not feed.
Why Is the Supply Chain Split Across So Many Countries?
Because each stage rewards something different, and firms located each stage where that thing was cheapest.
Malaysia is the clearest case, because the sequence is documented. The Free Trade Zone Act of 1971 and the Export Processing Zones established in 1972 offered duty-free imports of materials and capital equipment, ready infrastructure and tax holidays. American semiconductor firms arrived almost immediately — Intel, AMD, HP, National Semiconductor. By 1990, multinationals operating in those zones accounted for 85 per cent of total zone exports and 76 per cent of zone employment.
What they brought was the back end. Assembly and testing was, at that time, labour-intensive and low-capital: exactly what a country with an educated workforce and low wages could offer. What they kept elsewhere was design.
Every individual decision in that history was reasonable. A firm in California in 1974 wanting cheaper assembly, a Malaysian government wanting industrial employment, a Dutch company wanting packaging capacity near its Asian customers: none was deciding how the world’s supply of chips should be arranged. They were deciding where to put a building.
The aggregate is a production system in which no country makes a chip — and in which the number of places capable of performing any given step is far smaller than the number of firms depending on it. It is the same pattern this series keeps meeting, where an allocation nobody designed turns out to have distributed something that matters.
It is worth being careful here, because the obvious conclusion is not quite right. The instinctive response to the Nexperia stoppage is that Europe should make its own chips. But the wafers in that dispute were European, fabricated in the Netherlands, Germany and the United Kingdom. What Europe could not do at short notice was the cheap part — cut them up, package them, test them. Britain’s own parliamentary committee reached a version of this conclusion in 2022 and the government agreed: the country cannot and should not attempt to meet its semiconductor needs domestically. Onshoring the expensive, prestigious end of the chain would not have kept the Honda plant running.
Where Does the Money Go, and Where Does the Risk Sit?
In opposite directions. This is the finding.
Design captures the largest share of value in the industry while requiring the least physical capital — which is why the most profitable firms in the sector own no factories at all, and why a fabless company’s market capitalisation can exceed that of the foundries it depends on. Wafer fabrication takes a large further share. Assembly, testing and packaging, the stage employing the most people, takes the smallest.
I am stating that qualitatively rather than with a percentage, and the reason is worth putting in the open: the widely circulated figures for value distribution by stage trace back to a small number of subscription datasets, and I could not open a primary source for any of them. The direction is not in dispute and is visible in company accounts. The precise split is not something I can stand behind.
Now look at where a failure stops the world. Not in design: a design bottleneck delays next year’s product. Not, in this instance, in fabrication: the wafers existed. The stoppage happened at the back end, on mature devices, at the stage that captures least.
There is a reason, and it is not coincidence. Value accrues to whatever is hard to copy — accumulated intellectual property, process knowledge, patent estates. Fragility accrues to whatever is concentrated. Those are different properties. A stage can be easy to enter in principle, low-margin in practice, and still be concentrated in a handful of sites, because thirty years of cost competition drove everyone to the same few places. Cheapness and concentration are not opposites; cost competition produces concentration.
So the leverage in a global division of labour does not sit where the money sits. A country hosting the packaging of a commodity transistor has more capacity to halt European car production than a country hosting a chip design office — and earns a fraction as much for it.
What Happened Next Complicates This
The evidence here cuts against the neat version of the argument, and it should be said plainly rather than left out.
If the back end were genuinely irreplaceable, the story would have ended in November 2025. It did not. After the immediate crisis eased — the Busan summit produced a de-escalation, China lifted the export ban on 9 November, and the Dutch minister suspended his order ten days later as a gesture of goodwill — the two halves of the company did not reunite. They began replacing each other.
Nexperia’s Dutch headquarters had stopped shipping wafers to Dongguan at the end of October. Reporting through late 2025 and into 2026 indicates that the Chinese unit responded by qualifying domestic wafer suppliers, locking in local capacity to cover its 2026 production of one class of power chip, and declaring the Dongguan site its own global headquarters. The front end, in other words, turned out to be substitutable too.
But only slowly. Qualifying a new wafer supplier for automotive-grade parts is reported to take somewhere between two and three quarters, and during those quarters the Dongguan plant ran below capacity on depleting stocks of European silicon.
That is the more accurate version of the finding, and it is a better one. Leverage in a supply chain is not a matter of which stage is impossible to replace. Given enough time and enough state backing, most stages can be replaced. Leverage is a matter of how fast substitution is possible relative to how fast the damage accrues. The back end could stop European car plants within weeks. The front end could only starve the back end over quarters. Both were replaceable; only one could act at the speed of the harm.
The division of labour was organised to make production efficient. It ended up distributing the ability to interrupt production — and the ability to interrupt it quickly, which is the part that counts — in almost the reverse order to the ability to profit from it.
Who Loses in a Global Division of Labour?
Malaysia, and the countries positioned like it: the Philippines, Vietnam, and increasingly the states competing to succeed them.
The numbers first. Malaysia holds around 13 per cent of the global market in semiconductor packaging, assembly and testing, and is the sixth-largest semiconductor exporter in the world. The industry employs approximately 590,000 people. Semiconductors are the country’s largest export, at RM387 billion in 2022 — 21 per cent of total exports.
That is not a story of exploitation and should not be told as one. Survey evidence from the industry association reports that many of these are skilled jobs paying roughly double the national average. Malaysia’s position was won deliberately and it has been good for Malaysia.
The cost is of a different kind, and the evidence for it sits in the policy record rather than in a single figure. The Second Industrial Master Plan, from the mid-1990s, set out to move the country from basic assembly towards research and product development. The Third, from 2006, focused on higher value-added activities — R&D, design, engineering. The New Industrial Master Plan of 2023 aims to expand into high value-added activities within the value chain. Three decades of national planning, all pointed the same way. And a UNCTAD assessment published in 2025 still describes Malaysian semiconductor businesses as predominantly engaged in the mid-to-lower segments of the value chain, producing less complex, lower-specification products for foreign manufacturers and brand owners, with only a few players in the front end.
The mechanism holding a country in place is not tariffs and it is not a conspiracy. It is that the valuable stages are valuable because they are hard to enter: design depends on cumulative intellectual property and on engineers who learned the work at firms that already do it. The stage a country was invited into is the stage whose barriers to entry are lowest — which is also the stage from which the next country can undercut it.
And there is a sharper point in the numbers. Malaysia’s 13 per cent share sits overwhelmingly inside multinational facilities located in Malaysia. The plants are there. The process knowledge belongs to the firms that own them, and can go where the next incentive package is more attractive. A national share held in someone else’s factory is a share of the output, not of the capability.
Who avoided the cost? The firms and countries that kept design and intellectual property — not by taking anything from anyone, but by declining, decades ago, to relocate the one stage that compounds.
Britain is the mirror image, and worth a paragraph because the vulnerability is real but opposite. The UK’s strength is design and compound semiconductors, clustered around Cambridge, Bristol, South Wales and the North East. It has no volume fabrication of any scale. That earns well, and it means the country has no way to convert its position into supply when supply is short — as it was through the shortages that began in 2020, when British car plants stopped for want of components that British firms had helped design.
What Comparative Advantage Turned Into
The theory that organised all of this was about efficiency. Let each place do what it does relatively best, trade the results, and everyone ends up with more. As a description of how the semiconductor industry arranged itself, it is broadly accurate. The industry is spectacularly productive and chips are absurdly cheap.
What the theory did not describe — because it was not asking — is what a place’s position in the chain does to it politically. Specialisation does not only determine what a country earns. It determines what can be done to it, and what it can do to others, and how quickly.
A state hosting a mature packaging plant acquires a lever over car production on three continents. A state hosting a design cluster acquires income and no comparable lever. Neither acquired those things by seeking them. They fell out of decisions about where to put buildings, taken over fifty years by firms thinking about labour costs and tax holidays, cleared by governments who at the time were assessing an investment rather than a dependency.
That is what has changed, and it explains the strange tone of the last few years of industrial policy — the acts and strategies and consortia, all trying to buy back a position in a chain that no government arranged in the first place. Malaysia’s plans to climb it. Europe’s to reclaim the middle of it. Britain’s to defend the top of it. They are all late, and they are all responding to the same discovery: that a system built to be efficient turned out to be a distribution of power as well, and nobody was checking the second thing while they optimised the first.
More in this series
Previously: Undersea Cables
Next: AI as Infrastructure — a dependency forming faster than anyone can write down what has been installed.
All articles in this series: How Modern Society Works.
References
Named institutional analysis
Alicia García-Herrero, The Nexperia crisis: a wake-up call for Europe’s approach to Chinese investment, Bruegel, 23 December 2025. The principal source for the Nexperia sequence: Wingtech’s 2019 acquisition, Nexperia’s output and its share of European automotive supply, the US Entity List extension of 29 September 2025, the Dutch intervention of 30 September, the removal of the chief executive, China’s export halt of 4 October, and the November de-escalation.
UNCTAD, Towards a Green Semiconductor Industry in Malaysia, Project Paper No. 16, April 2025. Malaysia’s share of global packaging and testing, employment and export figures, the Free Trade Zone history, the sequence of industrial master plans, and the assessment of where Malaysian firms sit in the value chain.
Parliamentary
Business, Energy and Industrial Strategy Committee, report and government response on semiconductor supply, Session 2022–23. The UK’s design and compound semiconductor strengths, and the government’s position that domestic self-sufficiency is neither achievable nor desirable.