ASM International and ASMPT Smash Q2 Forecasts as AI Frenzy Fuels Chip-Tool Gold Rush

ASM International NV told investors on 21 April that second-quarter revenue will land around €980 million, comfortably ahead of the €930 million analysts had pencilled in. The Dutch company’s stronger projection follows a first-quarter print that also topped estimates, underscoring that the AI build-out is no longer a future tailwind but a present-day cash machine. Orders in the March quarter rose 30 % year-over-year to €815 million, the company said, confirming that every major logic and memory customer is racing to secure capacity for next-gen AI accelerators.

While AI garners headlines, China’s sudden snap-back in spending is amplifying the upside. ASM International noted that demand from Chinese fabs re-accelerated after a year-long lull, helping push total orders past the €800 million mark for the first time since late 2023. Beijing’s renewed appetite for mature-node capacity dovetails with global AI demand, creating a rare double-barrel growth story. The company cautioned that export-control clouds still linger, yet actual shipment data shows orders are converting into revenue with minimal friction.

Singapore-listed ASMPT Ltd., often viewed as the Asian cousin to ASM International, released its own bullish update only hours later. Management guided second-quarter revenue above consensus, citing “robust AI-related orders” for advanced packaging tools. Specifically, the company disclosed it has locked in equipment contracts for sixth-generation High-Bandwidth Memory (HBM4), the stacked DRAM technology that Nvidia, AMD and hyperscalers need to feed ever-larger AI models. Bookings for thermal-compression bonders used in HBM assembly are tracking 40 % above last year’s levels, according to supply-chain checks.

The synchronized beats from both ASMs reveal a supply-chain reality that Wall Street is only beginning to price: every new GPU or AI accelerator requires a cascading wave of wafer fab, deposition, lithography and packaging tools months before it reaches a server rack. When Nvidia announces a Blackwell Ultra or AMD teases a MI400, tool makers feel the spike almost immediately. That transmission mechanism has shortened planning cycles from quarters to weeks, leaving little slack in the system and pushing lead times for critical gear past six months.

For investors, the read-through is blunt: betting on AI winners no longer ends at Nvidia, AMD or Broadcom. The real torque sits upstream with tool vendors who sell the picks and shovels of the silicon gold rush. Foundries, meanwhile, face a dilemma. Securing ASM’s latest atomic-layer deposition chambers or ASMPT’s HBM bonders requires upfront cash commitments that can exceed $200 million per line, yet the payoff horizon is shrinking as AI product cycles accelerate. Those who lock in capacity now gain a two-year moat; those who hesitate risk missing the next wave entirely.

The exuberance carries caveats. US and Dutch export restrictions on advanced tools to China could cap up to 15 % of ASM International’s addressable market if geopolitical tensions flare again. On the memory side, HBM4 demand is predicated on continued AI-training growth; any sign that hyperscalers are over-provisioning GPUs would ripple back to equipment orders within a single quarter. Management teams at both companies admit they are running their factories at maximum capacity, leaving them vulnerable to any sudden demand air pocket.

Look for both companies to raise full-year numbers in July if current booking momentum holds. ASM International is already accelerating a €500 million capacity expansion in Almere, while ASMPT is scouting additional clean-room space in Penang and Suzhou. Customers including TSMC, Samsung and SK Hynix have quietly told suppliers they want delivery windows shortened to 20 weeks, a logistical stretch that could force smaller chip firms to the back of the queue and further consolidate AI supply chains around the top three foundries.