AMD said it will invest more than $10 billion in Taiwan, naming ASE, Powertech, Inventec and Sanmina as partners. The money is aimed not at front-end wafer production, but at advanced packaging capacity, a tighter constraint in the AI chip supply chain.
Advanced packaging moves to the center
In its May 21 announcement, AMD said the new investment will deepen cooperation with Taiwanese partners and expand packaging capacity. Advanced packaging refers to the process of integrating multiple chips onto a substrate at high density so they can communicate like a single processor package. Nvidia’s H100 and A100, as well as AMD’s MI300X, all rely on this manufacturing step, commonly associated with CoWoS.
Once wafers are finished, packaging becomes a critical gate before shipment. The source said the AI compute market was repeatedly constrained by CoWoS shortages from 2023 to 2025, and Nvidia also faced shipment delays tied to that bottleneck. In Taiwan, ASE and Powertech are identified as core suppliers in this part of the chain.
Lisa Su arrives in Taiwan as supply talks continue
AMD CEO Lisa Su has already arrived in Taiwan and is set to appear at a media fireside chat on Friday. According to the source, beyond attending Computex, she may also meet TSMC Chairman C.C. Wei and key supply-chain companies to seek access to TSMC’s 2-nanometer process capacity. That points to parallel efforts across both front-end manufacturing and back-end packaging.
AI spending is flowing toward packaging
The source also said large U.S. technology companies have accumulated more than $700 billion in AI spending for 2026. That capital ultimately moves through compute infrastructure, chips and packaging, leaving Taiwan in a central position. AMD’s public commitment to deepen its Taiwan ties puts fresh attention on a simple reality in AI hardware: shipment capacity depends not only on chip design, but also on who can secure packaging lines.