ASML CEO Christophe Fouquet: Nobody is coming for us

Each time you employ AI, you might be, in some small means, relying on a 42-year-old, 44,000-person Dutch firm that spends €4.5 billion annually to advance its know-how.

ASML, headquartered within the Netherlands, makes the machines that make the chips that make AI potential. Extra particularly, it makes the one machines on the earth able to printing the microscopic patterns on silicon wafers that outline probably the most superior semiconductors — a course of referred to as excessive ultraviolet lithography, or EUV. The machines are roughly the dimensions of a faculty bus, take months to assemble, contain a whole bunch of suppliers, and value wherever from $200 million to upwards of $400 million apiece relying on the era (costs that give even ASML’s greatest prospects pause sometimes).

That monopoly has made ASML probably the most beneficial firm in Europe, price over $530 billion. And with the 4 largest American tech corporations — Microsoft, Meta, Amazon, and Google — committing greater than $600 billion in AI infrastructure spending this 12 months alone, demand for ASML’s machines has surged to the purpose the place the corporate has overtly mentioned the world gained’t have sufficient chips for years.

All that demand has additionally made ASML a goal. Substrate, a San Francisco startup based by a protégé of Peter Thiel, has raised greater than $100 million and been valued at over $1 billion on the declare that it might construct a rival lithography machine. Individually, there have been reviews that former ASML engineers in China have partly reverse-engineered the know-how, a prospect with huge geopolitical implications.

Christophe Fouquet, who turned ASML’s CEO in 2024 after greater than a decade on the firm, sat down with this editor on the rooftop deck of his Beverly Hills lodge Tuesday morning forward of his look on the Milken Institute World Convention. Wearing a blue swimsuit and white shirt, he was relaxed — even when the dialog turned to the rivals.

This interview has been evenly edited for size and readability.

Did you see the AI explosion coming?

No, by no means. We labored very exhausting, however not with the concept that this is able to come. You went from an idea — one thing individuals thought would finally arrive — to ChatGPT, which was actually the primary good instance of what AI may do. And now I feel we have a look at AI as the subsequent revolution, not solely industrial however societal. Did I see it coming? No. Sitting in the midst of it daily, typically we get up within the morning and nonetheless test that what is occurring is absolutely taking place.

The massive query everybody has is whether or not the availability chain can hold tempo with demand. Can it?

The demand is such that the market total can be supply-limited for fairly a bit. Proper now, the largest bottleneck appears to be in chip manufacturing. We, as an gear provider, comply with our prospects, and to date we’ve adopted them fairly properly — however we all know we’ve to step up our whole provide chain and capability. If you happen to discuss to the hyperscalers, I feel they’ll inform you that for the subsequent two, three, even 5 years, they’re not going to get sufficient chips.

TSMC made information not too long ago saying your newest machines are too costly. How do you reply?

An EUV system, should you have a look at the worth, goes to be dearer than a low-NA system, however the price of making a wafer with this instrument on some superior layers can be cheaper. We are able to get 20%, 30% value discount.

[Editor’s note: Both machines Fouquet is referring to here are EUV machines — the same fundamental technology. NA stands for numerical aperture, a measure of how finely a machine can focus light onto a chip. Low-NA EUV is the current generation; high-NA EUV is ASML’s newest generation, capable of printing even finer patterns but carrying a price tag of $350 million or more apiece. Fouquet is arguing that even though the new machine costs more, it produces chips more cheaply.]

I get plenty of questions on whether or not it’s going to be this month or subsequent month or the month after. And I normally say it doesn’t actually matter, as a result of we designed high-NA for the subsequent 10, 20 years. You may return to the press from 2016, 2017, and also you’ll discover the identical quotes — low-NA EUV was very dear. We all know what occurred after that. The identical will occur with high-NA.

There’s a startup referred to as Substrate, backed by Peter Thiel, claiming it might construct a rival lithography machine. What do you consider it?

Eager to have it and having it — that’s nonetheless an enormous distinction. The challenges of lithography are many. Having the ability to make a picture is a place to begin, however that you must make that picture in very excessive amount, at very low value, at excessive pace, and with nanometer accuracy. I all the time say the one purpose ASML may construct an EUV machine is as a result of 80% of it already existed, primarily based on earlier data and merchandise constructed over time. We needed to resolve one downside — getting EUV mild — and that alone took 20 years. Whenever you begin from scratch, the problem is gigantic. I’ve seen plenty of claims. I’ve seen a couple of footage. However we had our first EUV image 30 years in the past, and we nonetheless wanted 20 extra years of exhausting work to show it into a producing system.

What about xLight, a laser startup partly backed by the U.S. authorities that wishes to work with you?

xLight is specializing in one aspect of our EUV machine — the supply that creates the sunshine. The supply we’ve might be prolonged for a few years to return, and we all know tips on how to scale it. What xLight is doing is a brand new supply that also must be constructed and confirmed. The one query is whether or not it gives a efficiency or value benefit over what we’ve. I feel the jury continues to be out. We’re working with them to allow them to display their know-how — we really feel that’s a duty on our aspect. But it surely’s nonetheless a really lengthy journey.

There are additionally reviews that former ASML engineers in China have reverse-engineered your machines.

To reverse-engineer something, you first have to have the machine. And there’s no EUV machine in China — we by no means shipped any instruments there. All of the instruments we’ve shipped, we all know the place they’re. They’re both in use with prospects, and we observe these, or they’ve been dismantled and got here again to us. The concept one in every of our programs is in China is solely unsuitable. And since our EUV know-how has by no means been exported there, we additionally don’t have any individuals in China skilled on EUV.

Very early on, when restrictions got here in, we created a whole separation inside the firm between those that can entry EUV know-how, paperwork, and coaching, and those that can’t. Our crew in China sits on the opposite aspect of that line. The information level to little or no, if any, progress in any respect. It’s exhausting for individuals to simply accept that as a result of entry to this know-how is so vital.

On export controls extra broadly — Jensen Huang was right here final evening arguing that corporations ought to promote globally, that extra company income means extra tax {dollars} for a corporation’s house nation. He additionally mentioned the vital factor is to maintain one of the best and newest nearer to house. Do you agree?

I feel he’s completely proper. What he provides — and I feel that is what Nvidia has executed — is which you could hold a technological benefit by sustaining a era hole in what you promote. Nvidia sells a couple of generations again, and that lets them discover the steadiness between nonetheless doing enterprise and never handing a robust aggressive benefit to nations the place you gained’t promote the newest. We imagine the identical strategy ought to apply to our merchandise. At present we ship instruments to China — allowed by export controls — but it surely’s a instrument we first shipped in 2015. If you happen to apply Jensen’s philosophy to our scenario, Nvidia is working with roughly an eight-generation hole. We’re two or three. There’s room for rationalization — discovering the precise steadiness between not doing enterprise in any respect, shedding a significant alternative, and strongly inviting others to compete with you.

How do you assess the place issues stand with the present administration on all of this?

There’s a good dialogue, which is essential. I feel there’s a real understanding of what enterprise wants, however there’s nonetheless the problem of discovering the precise steadiness between all of the completely different voices and pursuits. The dialogue is there, and we respect that. I’ve been in Washington many instances. A minimum of the dialogue is occurring. But it surely’s a really complicated matter.

You don’t appear involved about anybody short-cutting your know-how.

Folks prefer to have the best know-how, however they have a tendency to neglect what it took to construct it. It’s been a few years of labor — not solely at ASML however with our suppliers. Many alternative teams of individuals fixing very tough issues, after which one firm bringing all of it collectively utilizing a long time of lithography experience to show it into a producing system. That is by no means straightforward. And I feel that’s additionally our greatest safety. It’s merely what it took to place it collectively.