ASML customers are enquiring about machines for 2030, exposing the production bottleneck behind the technology industry’s increasingly extravagant AI spending promises.
Financial advisory and fintech outfit deVere Group chief executive Nigel Green warned that investment pledges cannot conjure equipment into existence. His comments follow ASML’s announcement that customers are expressing interest in placing orders for 2029 and even 2030.
The Dutch manufacturer is exploring ways to expand 2028 capacity beyond the 30 per cent increase previously communicated. Its currently planned supply for 2027 and 2028 will fall short of customer demand, according to ASML chief executive Christophe Fouquet.
“Everyone’s focused on how much money is pouring into AI. The more useful question is how fast it can actually be spent, and that comes down to how many of these machines get built each year,” Green said.
“A trillion-dollar pledge doesn’t mean much if the equipment won’t turn up until 2030,” Green said.
ASML remains the sole commercial supplier of extreme ultraviolet lithography systems used to manufacture the most advanced chips. That puts the company at the centre of chipmakers’ expansion plans for the AI industry.
Green said: “Only one company in the world can make the machines the most advanced chips depend on, and there’s a limit to how many it can build. You can raise all the capital you like, but if the equipment isn’t there, the money sits and waits.”
Green said equipment enquiries stretching four years ahead illustrate constraints beyond ASML, including electricity connections, electrical equipment, and specialist workers. Those shortages affect data centre construction and operation, giving spending announcements several chances to collide with reality.
“When a buyer is asking about machines for 2030, they’re telling you how long the queue is. Some of this year’s AI plans will get stretched out, and some will quietly shrink,” he said.
He argued that government and corporate pledges depend on equipment, power and production slots that have yet to be secured. ASML raised the lower end of its 2026 sales forecast, with Green interpreting continued demand as evidence of bottleneck suppliers’ pricing power.
“Money can buy a lot in this market, but it can’t buy a machine that hasn’t been built yet. Companies that locked in their slots early have an edge their rivals will struggle to close, no matter how much they spend. Every AI plan in the world now runs to the same clock, and writing a bigger cheque won’t make it tick any faster,” Green said.
