AMD unveils US$10 billion Taiwan AI investment, ramps up TSMC 2nm chip production

Taipei, May 21 (CNA) U.S. semiconductor company AMD will invest more than US$10 billion in Taiwan's AI supply chain ecosystem while ramping up production of its next-generation server processors using Taiwan Semiconductor Manufacturing Co's (TSMC) advanced 2-nanometer process technology.

In a press statement issued Thursday, AMD said the investments will expand partnerships with Taiwanese chip packaging, substrate and server manufacturing companies as global demand for AI infrastructure continues to surge.

AMD Chair and CEO Lisa Su (蘇姿丰) said the company is combining its high-performance computing technologies with Taiwan's manufacturing ecosystem and global partners to build integrated, rack-scale AI infrastructure for next-generation AI systems.

"By combining AMD leadership in high-performance computing with the Taiwan ecosystem and our strategic global partners, we are enabling integrated, rack-scale AI infrastructure," Su was cited as saying.

A major focus of the investment plan is advanced packaging technology for AI chips, including next-generation 2.5D bridge interconnect technology known as Elevated Fanout Bridge (EFB), according to AMD.

To advance the technology, AMD said it is collaborating with Taiwan-based ASE Technology Holding Co. and Siliconware Precision Industries Co. (SPIL) to develop EFB packaging capable of improving chip interconnect bandwidth and power efficiency for future EPYC processors.

AMD also said it had achieved a milestone with Powertech Technology Inc. (PTI) by qualifying the industry's first 2.5D panel-based EFB interconnect, which it described as enabling more scalable and cost-efficient AI chip production.

The investments are also tied to AMD's upcoming Helios AI platform, which the company described as a production-ready rack-scale AI infrastructure system scheduled for deployment beginning in the second half of 2026.

In another statement issued the same day, AMD announced that production of its next-generation EPYC server processor, codenamed "Venice," has been ramped up in Taiwan using TSMC's advanced 2nm process technology, making it the first high-performance computing product in the industry to do so.

Growing "agentic AI" workloads are increasing the importance of CPUs in coordinating data movement, networking, storage and system orchestration across AI data centers, AMD said.

Future production of "Venice" processors is also planned at TSMC's fabrication facility in Arizona as AMD seeks to diversify its advanced manufacturing footprint, the company added.