AMD announced a US$10 billion [1] investment to expand Taiwan’s AI chip ecosystem and increase manufacturing capacity.
This move aims to secure the critical infrastructure needed to meet surging global demand for AI chips. By deepening its partnership with TSMC, AMD seeks to reduce its reliance on specific packaging technologies and ensure a stable supply chain for its high-end hardware.
CEO Lisa Su arrived in Taiwan on Friday, May 22, 2024, to finalize the details of the agreement. During her visit, Su attended a forum in Taipei and held meetings with TSMC Chair C.C. Wei [2, 3]. The investment, described by some sources as over US$10 billion [4], will focus heavily on advanced packaging and manufacturing capacity.
Securing this capacity is vital as the industry moves toward smaller, more efficient transistors. AMD is targeting the two nm manufacturing process [1] for its next-generation "Venice" chips. The shift to this advanced node requires significant infrastructure upgrades and a tightly integrated relationship with the foundry that produces the silicon.
Central to the strategy is the goal of reducing reliance on CoWoS, the Chip-on-Wafer-on-Substrate packaging method that has historically created bottlenecks for AI chip producers [3, 5]. The new investment is intended to diversify and scale these packaging capabilities within Taiwan to prevent future production delays.
The collaboration underscores the interdependence between U.S. chip designers and Taiwanese fabrication. By investing directly into the ecosystem, AMD is attempting to insulate its roadmap from the volatility of global semiconductor logistics, a necessity as AI integration accelerates across all computing sectors.
“AMD announced a US$10 billion investment to expand Taiwan’s AI chip ecosystem.”
This investment signals a strategic shift toward vertical integration of the supply chain. By funding the infrastructure in Taiwan, AMD is not just buying chips but securing the physical means of production. This reduces the risk of supply shortages that have plagued the AI sector and positions the company to compete more aggressively with Nvidia by ensuring its 2 nm hardware can actually be packaged and shipped at scale.
