When One Company’s Restriction Becomes Another’s Ceiling
Nvidia’s tightening grip on AI chip exports – driven by U.S. government restrictions targeting China and other designated markets – is doing something quietly damaging to the broader semiconductor supply chain. It is shrinking the addressable market for the machines that make those chips possible. ASML, the Dutch lithography giant whose extreme ultraviolet machines are the only way to manufacture cutting-edge chips at scale, finds itself caught in a constraint it did not create and cannot easily escape.
The connection is not obvious at first glance, but it is direct.
When Nvidia cannot sell H100s or Blackwell GPUs into a market, demand for next-generation chip fabrication capacity drops in that same market. TSMC does not build new fabs speculatively. Samsung does not order additional EUV machines on a hunch. Capacity expansion follows confirmed demand, and if that demand is blocked at the sales level, the equipment orders never materialize. ASML sits at the very end of that chain, which means it absorbs the impact last – but also most quietly, in ways that don’t generate the same headlines as Nvidia’s export filings.
ASML’s China Problem Isn’t New, But It’s Getting Worse
ASML has been restricted from shipping its most advanced EUV systems to China since 2019, when the Dutch government – under pressure from Washington – declined to renew ASML’s export license for EUV tools. That decision alone cut off what would have been a major revenue stream. China’s chip ambitions are enormous, and its willingness to pay for advanced equipment has never been in question. The restriction was a political decision with permanent financial consequences for ASML’s long-term growth ceiling in that region.
What the Nvidia export curbs add to this picture is a compounding layer. Even older deep ultraviolet lithography machines, which ASML was still permitted to sell into China under certain conditions, are now caught in a tightening web. The U.S. has pressed allies to restrict more categories of chip-related equipment, and the Dutch government has responded with additional licensing requirements on DUV tools. China accounted for a significant portion of ASML’s revenue in recent years – at certain points representing close to half of its total quarterly shipments by unit. Closing that channel further, even partially, is not a footnote in ASML’s earnings. It is a structural revenue question.
ASML’s order backlog has remained strong on paper, buoyed by orders from TSMC, Intel, and Samsung as those companies race to build out capacity for AI workloads. But backlogs reflect commitments made under assumptions that can shift. If Nvidia’s export restrictions slow AI chip demand in key growth markets, the long-term case for aggressive fab expansion weakens – and with it, the case for additional EUV machine orders beyond what is already contracted.
The Quiet Squeeze on Future Demand
The most immediate financial reading on ASML still looks healthy. The company’s EUV machines sell for roughly 200 million euros each, and its high-NA EUV systems – the next generation of the technology – are priced even higher, with only a handful of buyers capable of purchasing them at all. Revenue per tool is extraordinary. But the business model depends on a steady expansion of who is buying and how many tools they need. A world where AI chip production is geographically fragmented by export controls is a world where the total number of advanced fabs – and therefore the total number of machines those fabs require – grows more slowly than it otherwise would.
There is also a demand-signal problem that plays out over a longer horizon. Nvidia’s restrictions push some buyers toward less advanced chips or toward domestically produced alternatives in markets like China, where companies such as Huawei have been developing their own AI chip architectures. Those chips are manufactured on older nodes using equipment ASML has already sold, not new EUV orders. Every design win for a domestic Chinese chip running on mature-node SMIC capacity is, by definition, a design win that does not require a new ASML shipment.
The irony is that the U.S. export control strategy was partly designed to prevent China from accessing exactly the kind of advanced manufacturing capability that ASML’s tools represent. It has succeeded in that narrow goal. But the side effect is that it has also capped the demand pool that would otherwise absorb ASML’s production capacity for years to come. The company is not in financial distress – its position in the global chip ecosystem is genuinely without parallel, since no other company produces EUV machines at commercial scale. But “without parallel” is not the same as “without limits.”
What ASML Cannot Control
ASML’s leadership has consistently flagged geopolitical risk as a key variable in its long-term guidance, and the company has made reasonable efforts to diversify its customer base across U.S., European, Japanese, and South Korean chipmakers. None of that changes the underlying arithmetic: a market representing a substantial share of global semiconductor demand has been progressively walled off, and the walls are being built taller every time Washington tightens the Nvidia export framework. ASML does not set trade policy, does not control where Nvidia’s chips are permitted to go, and cannot manufacture demand for EUV tools in markets that have been restricted from accessing the end products those tools would ultimately help produce. The company is exceptional at what it does – and entirely dependent on a geopolitical environment it has no power to negotiate.
