Advanced Micro Devices is working with partners in Taiwan to increase production capacity as stronger-than-expected demand tightens the global market for central processing units, Chief Executive Lisa Su said in Taipei.
Su said demand for CPUs has moved above what the company expected a year ago, adding pressure to a market that had been relatively steady before the latest wave of artificial intelligence infrastructure spending. AMD expects supply to increase every quarter this year, with larger increases planned in 2027 and beyond, Reuters reported.
The comments came as AMD deepens its manufacturing and packaging ties in Taiwan, a central hub in the global semiconductor supply chain. Taiwan is home to TSMC, AMD’s key foundry partner and the world’s largest contract chipmaker, as well as companies that handle advanced packaging, substrates, testing and server assembly.
AMD said this week it will invest more than $10 billion across Taiwan’s technology ecosystem to expand strategic partnerships and scale advanced packaging manufacturing for next-generation AI infrastructure. The company said the investment is aimed at supporting rising demand for AI systems and improving its ability to deliver rack-scale platforms built around its CPUs and accelerators.
The company is also ramping production of its next-generation EPYC processor, codenamed “Venice,” using TSMC’s 2-nanometer process technology. AMD said Venice is its sixth-generation EPYC server CPU and the first high-performance computing product to begin a production ramp on TSMC’s advanced 2nm technology.
The capacity push reflects a change in how the AI market is affecting chip demand. Much of the early AI infrastructure boom was centered on graphics processing units, which are used to train and run large AI models. But CPUs remain essential in data centers, coordinating workloads, feeding accelerators, managing memory and handling parts of AI systems that do not run entirely on GPUs.
Su said growth is being driven by AI inferencing and agentic AI, Reuters reported. Inferencing is the process of running AI models after they have been trained, while agentic AI refers to systems designed to carry out tasks more independently. Those workloads can increase demand for both GPUs and CPUs, particularly as companies deploy AI services at scale.
AMD’s Taiwan expansion also covers advanced packaging, a process that connects multiple chip components into a single high-performance system. That work has become increasingly important as chipmakers combine CPUs, GPUs, memory and other components to improve speed and power efficiency. AMD said its Taiwan investment includes packaging technologies for Venice CPUs and its Helios rack-scale platform, which is designed to pair Venice with AMD Instinct MI450X GPUs.
Reuters reported that AMD is working with Taiwanese partners including ASE, SPIL, PTI, Wiwynn, Wistron, Inventec, Unimicron, AIC, Nan Ya PCB and Kinsus. Su said the company is co-investing with partners to secure enough capacity for expansion in 2026 and beyond, including through 2029.
The push comes as AMD competes with Nvidia in AI infrastructure while also seeking to strengthen its position in server CPUs. Nvidia remains dominant in AI accelerators, but AMD is betting that demand for complete AI systems will create more room for its processors, accelerators and rack-level designs.
Taiwan’s role in that strategy is significant. TSMC manufactures advanced chips for many of the world’s largest technology companies, and Taiwan-based suppliers are deeply involved in packaging and assembling the systems used in data centers. AMD’s investment signals that the company expects demand for AI infrastructure to remain strong enough to justify long-term capacity commitments.
AMD is also balancing its global growth plans with export-control limits. Su said China remains an important market for the company and accounts for about 20% of AMD’s revenue, Reuters reported. She said AMD will continue working with Chinese customers while complying with U.S. restrictions on certain high-end AI chip shipments.
For data center customers, the capacity ramp could help ease supply pressure in a market where AI demand has strained availability for advanced chips and related components. For AMD, it is a bet that CPUs will remain central to AI infrastructure even as GPUs continue to draw most investor and customer attention.
The company’s latest announcements point to a longer-term buildout rather than a short-term production adjustment. AMD is positioning Taiwan as a core part of its AI supply chain, with new investment tied to Venice CPUs, advanced packaging and large-scale AI systems expected to roll out through the second half of 2026 and beyond.