The AI infrastructure boom shows little sign of easing

One of the clearest ways to gauge the pace of the AI boom is by watching the hardware supply chain. NVIDIA is often the most visible indicator: as AI companies race to build massive data centres, they are purchasing GPUs at an unprecedented rate, helping propel NVIDIA to become the world’s most valuable company. But Nvidia itself depends on upstream suppliers — and those suppliers can offer an even longer-term view of where the market is headed.

That’s where ASML comes in. The Netherlands-based photolithography firm plays a uniquely critical role in the global semiconductor ecosystem. ASML is the only company in the world capable of producing extreme ultraviolet (EUV) lithography machines, which are required to manufacture the most advanced chips. As a result, ASML sits at the centre of nearly every major semiconductor roadmap. When demand at ASML accelerates, it usually signals that chipmakers expect strong sales well into the future.

Based on the company’s quarterly earnings released Wednesday morning, the demand is accelerating sharply.

ASML reported €32.7 billion in net sales — a substantial figure by any measure. But the more telling metric is “new bookings,” which reflect fresh orders placed during the quarter. Those bookings provide insight into how much production capacity chip manufacturers believe they’ll need, based on anticipated demand from AI-driven data centre expansion in the years ahead.

On that front, the numbers are striking. During the most recent quarter, ASML recorded €13 billion in new orders — the highest quarterly total in the company’s history and more than double the bookings from the prior quarter.

In an earnings statement, ASML CEO Christophe Fouquet made clear that artificial intelligence is the driving force behind the surge.

“Over the past months, many of our customers have expressed a significantly more positive view of the medium-term market outlook, largely due to stronger confidence in the durability of AI-related demand,” Fouquet said. Translated from executive language, that means chipmakers believe AI labs will need every bit of the data centre capacity currently under construction — and they’re spending aggressively now to be ready when that demand materialises.

Of course, none of this is guaranteed. These machines take years to deliver, and some orders could still be delayed or cancelled before completion. Sceptics continue to warn that overly optimistic projections could unravel if AI spending cools or expectations fail to materialise.

Still, for anyone expecting companies to start retreating from the trillions of dollars earmarked for AI infrastructure, the latest data suggests that moment hasn’t arrived — and may not for some time.