Chip Manufacturing

ASML's Strong Orders Calm Fears AI Chip Capex Boom Is Cooling

ASML's order book held strong enough to ease fears the AI chip spending cycle is losing steam, the Korea Times reported this week, framing the EUV monopolist's bookings as a fresh read on the AI cycle.

By
Rebecca Stone
Filed
Channel
Chip Manufacturing
Read
3 min read

ASML's order book came in strong enough to push back fears that the AI infrastructure spending cycle is running out of steam, the Korea Times reported this week. The headline matters because ASML sells into nearly every leading-edge fab on the planet.

Its extreme ultraviolet, or EUV, lithography systems remain the only tools on Earth capable of patterning the sub-7nm layers inside today's AI accelerators and high-end CPUs. The company sits at a near-monopoly chokepoint for advanced chipmaking.

When hyperscalers order more silicon, the orders cascade backward through foundries and memory makers — eventually landing on ASML's books months before any of that new capacity goes live.

Why does ASML's order book move the AI narrative?

When TSMC, Intel, Samsung Foundry, or SK Hynix commit to a new fab or a new process node, ASML gets the order months before the equipment ships. That pipeline makes its quarterly intake the cleanest forward-looking signal of how much new chip capacity the industry intends to add.

Strong ASML orders don't just track foundry capex. They pull in memory makers too.

SK Hynix, Samsung Electronics, and Micron have all been expanding HBM3 and HBM3E capacity to feed AI accelerator programs, and that build-out relies on EUV-class patterning for the underlying DRAM and logic die. Strong bookings, in other words, mean customers across both logic and memory are still writing checks for capacity that won't produce revenue for two to three years.

What was the AI capex worry?

The Korea Times report landed just as analysts had begun asking whether the AI capex surge had already peaked. Demand for high-bandwidth memory and cutting-edge logic had pulled forward massive investment across memory and foundry capacity.

Some on Wall Street had started wondering how much runway remained, especially as the early NVIDIA GB200 ramp gave way to more measured guidance from the AI supply chain.

The Korea Times framing — that ASML's bookings ease those worries — hints that the order intake didn't fall off the cliff some bearish forecasts had predicted. Whether that translates into sustained revenue growth through 2025 and 2026 still depends on the pace of node migration and the timing of new fab ramps in places like Arizona, Kumamoto, and Yongin.

What risks could still turn the cycle?

The geopolitical backdrop muddies the read. The US government's export controls on advanced EUV systems to China have reshaped ASML's customer mix, and any further tightening — or any thaw — would ripple through the backlog.

The Dutch government, working alongside Washington, has tightened licensing on EUV shipments to Chinese fabs since 2023. ASML's China sales have already swung sharply lower as a share of total revenue, and the company's installed base there remains under watch.

Memory pricing volatility adds more noise. One strong quarter is one strong quarter.

Bears will keep watching China exposure, the pace of HBM3E qualification, and whether the next round of hyperscaler AI revenue justifies the next round of fab capex.

What does this signal going forward?

ASML's strong orders, as the Korea Times reported, point to a chip capex cycle that still has runway through the next two production ramps.

The real read on whether the cycle holds will come with ASML's full-year 2025 bookings, when the company reports how much of its backlog converts into shipped tools, and whether the next generation of 2nm-class process nodes — TSMC's N2, Samsung's 2nm, and Intel's 18A — hit their planned ramps.

Source: Google News: AI chips

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

467 articles

Related articles

« Previous article