AMD CEO Lisa Su said the company plans to further increase investment in Taiwan's semiconductor supply chain as demand for CPUs, GPUs, and AI computing continues to outpace available capacity.
Speaking to reporters during a brief visit to Taiwan on October 6, 2026, Su said AMD expects to significantly increase product supply in 2027, but the market will still be able to absorb additional capacity. The company is therefore working with suppliers on longer-term capacity planning, extending beyond the traditional one- to two-year horizon to three, four, or even five years.
Su said AMD has maintained close relationships with Taiwan-based partners across wafer manufacturing, advanced packaging, and substrates. She also highlighted the importance of geographic diversification while stressing that Taiwan remains "absolutely critical" to both the global semiconductor supply chain and AMD.
AMD's next-generation Venice CPU, built using TSMC's 2nm process, has already achieved strong adoption among major hyperscalers, cloud providers, and enterprise customers, Su said. Despite tight 2nm capacity, she expressed confidence that AMD and TSMC will be able to supply market demand.
Advanced packaging is also becoming increasingly important for high-performance computing. Su said AMD is exploring packaging innovations across its supply chain and ways to add capacity, while continuing its close partnership with TSMC.
Su also confirmed that AMD's Helios AI platform began shipping as planned in the third quarter of 2026 and will continue ramping over the coming quarters as demand for computing capacity remains strong.
AMD announced more than US$10 billion in investments in Taiwan's supply chain during Su's May visit. She said the investment remains on schedule and will be increased further as market demand continues to rise. Su added that AMD's existing business model of working with major global suppliers remains effective, rather than following moves by some competitors to directly invest in chip manufacturing.
On AI development, Su rejected calls to slow innovation, arguing that rapid technological progress and safety can be pursued simultaneously. She said the industry must work collectively to develop highly capable AI while addressing potential negative effects and safety concerns.
Following AMD's agreement to acquire World Labs, Su said the company is pursuing innovation across hardware, software, systems, and models. She expects closer interaction between hardware and software as AI models evolve.
Su also said AI will develop across enterprise, research, and physical-world applications, with different technologies serving different use cases and potentially producing "many winners" in the market. She further welcomed greater coordination among leading AI companies through the US government's new superintelligence initiative.
Article edited by Ysi Chen
