Austrian circuit-board and substrate maker AT&S will invest up to €2 billion in its Malaysia operations to expand capacity for advanced chip packaging tied to AI accelerator demand. The commitment, disclosed by the company, marks one of the largest single buildouts by a European supplier in Southeast Asia's semiconductor corridor and locks AT&S into the upstream supply chain feeding Nvidia-class GPUs and custom AI silicon.
The €2 billion figure represents an upper-bound spend across a multi-year horizon at AT&S's Kulim site in Malaysia, where the company already manufactures IC substrates — the high-density boards that connect a processor die to the rest of a system. Substrates have become a bottleneck in the AI hardware chain, with advanced packaging supply constraining how many accelerators hyperscalers can deploy each quarter.
AT&S is pitching the Malaysia expansion as a direct response to AI customer demand. The company supplies substrates used in high-end compute packages, a category dominated by a small number of producers including Japan's Ibiden and Shinko, Taiwan's Unimicron, and South Korea's Samsung Electro-Mechanics. European participation in this tier has historically been thin, and AT&S has positioned its Kulim facility as the West's hedge against East Asian concentration.
Key facts
- 01AT&S will invest up to €2 billion in its Malaysia operations to expand advanced substrate capacity.
- 02The investment is explicitly tied to AI chip demand and high-end packaging for accelerator customers.
- 03Malaysia is emerging as a core node in the AI semiconductor supply chain alongside Taiwan and South Korea.
Malaysia's role in the AI supply chain has accelerated quickly. The country has long anchored back-end semiconductor assembly and test, and is now climbing into higher-value packaging steps as TSMC, Intel, and substrate suppliers stand up regional capacity. Kulim, in the northern state of Kedah, sits at the center of that buildout. AT&S's commitment cements the location as a strategic node alongside Penang and Johor.
The financial stakes are substantial for a company AT&S's size. The Leoben-based group's full-year revenue runs in the low single-digit billions of euros, meaning a €2 billion capex commitment to a single geography represents a multi-year bet that AI accelerator volumes will keep absorbing advanced substrate output through the end of the decade. The company has already booked multi-year supply agreements with unnamed AI chip customers to underwrite the spend.
AT&S has faced execution challenges at Kulim before. Earlier ramp phases at the Malaysia plant ran into delays and cost overruns that pressured the company's margins and forced management changes. The new commitment effectively doubles down on the site rather than diversifying, betting that demand visibility from AI customers now justifies absorbing the operational risk.
The competitive context favors aggressive capacity expansion. Nvidia, AMD, and the hyperscaler in-house silicon teams at Google, Amazon, and Microsoft are all increasing accelerator order volumes for 2026 and 2027 deliveries. Advanced substrates and CoWoS-class packaging remain the most-cited supply constraints in those builds, and any supplier that can credibly ship at scale into that window stands to capture pricing power that didn't exist five years ago.
Geographic placement also matters for customer routing. Locating substrate production in Malaysia puts AT&S geographically adjacent to the assembly, test, and packaging steps that increasingly happen outside Taiwan as customers diversify post-2022 supply-chain shocks. Shorter logistics paths between substrate manufacture and final accelerator packaging cut both lead time and inventory cost for AI hardware buyers.
Risks to the thesis are real. AI accelerator demand has been forecast aggressively before, and substrate suppliers that built capacity into the 2022 cycle were left with excess inventory when crypto and consumer GPU demand collapsed. A slowdown in hyperscaler capex, a shift in packaging architecture, or a faster-than-expected ramp by larger Asian competitors could all leave AT&S overbuilt. The company's history of execution slippage at Kulim raises the bar for delivering this expansion on time and on budget.
The bigger signal from this commitment is what it says about where AI value is concentrating. The trillion-dollar narrative around AI sits at the model layer with OpenAI, Anthropic, and Google, but the binding constraint on the industry's growth keeps showing up in unglamorous places: substrates, advanced packaging, HBM, power delivery, and cooling. Suppliers that can credibly add capacity in those layers are quietly capturing structural pricing power that the model labs themselves don't have. AT&S's €2 billion bet is a wager that the constraint stays binding long enough to earn the capex back, and that European industrial firms can hold a defensible seat at a table that has, until now, been set almost entirely in East Asia.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
