AMD bets $10b on Taiwan AI expansion

US chipmaker AMD said on May 21 it will invest more than US$10 billion in Taiwan’s AI sector to expand its capacity to build and assemble advanced AI chips.

The company also plans to deepen partnerships there.

AMD said it will work with Taiwan’s ASE and its unit SPIL on more power-efficient technology for AI systems and processors.

It will support its Venice CPUs made on TSMC’s 2 nanometer process.

The company also said it has started ramping production of Venice CPUs and is working with other Taiwanese partners including PTI, Sanmina, Wiwynn, Wistron, and Inventec.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

AMD’s Taiwan push also rides a CPU return in AI

AMD is centering its Venice central processing units (CPUs) on a fresh wave of data center demand, after graphics processing units (GPUs) soaked up much of AI spending 1.
That change is coming from large tech groups such as Meta, which is rolling out CPU-only servers for some new AI tasks and has signed deals with AMD for Venice chips 1.
The move ties the U.S. more closely to Taiwan’s chip industry, which already makes about 90% of the most advanced logic chips used in AI 2.
It comes while the U.S. government presses to bring home 40% of advanced chip supply, a target some Taiwanese officials have called “impossible” 2.

The investment exposes weak spots in the AI chip supply chain

Closer work with Taiwan leaves AMD more exposed to supply bottlenecks that reach beyond military risk in the region 2.
Advanced chipmaking relies on minerals such as germanium, yet Chinese shipments of that material to Taiwan dropped to zero from 2022 to 2024 2.
A helium shortage, worsened by the war in Iran and a production stop in Qatar, also puts pressure on advanced semiconductor manufacturing 2.
This concentration may push more countries to build local chip industries, even with the huge cost and difficulty of copying Taiwan’s system 2.