According to tspasemiconductor.substack.com, AMD is shifting from one- or two-year procurement cycles to three- to five-year capacity planning across Taiwan’s semiconductor supply chain — a strategic pivot aligned with surging demand for EPYC CPUs and Instinct GPUs.
Strategic Shift in Capacity Planning
Lisa Su, AMD’s chief executive, led a high-priority visit to Taiwan on October 6, meeting key partners including TSMC, Foxconn, Quanta, Asus, and Acer. The itinerary spanned advanced chip manufacturing, server assembly, and PC branding — effectively mapping Taiwan’s full computing supply chain. This was not a ceremonial tour but a coordinated negotiation to secure long-term capacity, as Su confirmed AMD increased supply during 2026 and expects to increase it “substantially” again in 2027.
$10 Billion Ecosystem Commitment Underway
AMD reaffirmed that its more than $10bn commitment to Taiwan’s technology ecosystem — first announced in May — is proceeding as planned. The company stated the amount will increase further as demand rises, though no revised figure has yet been disclosed. This investment supports infrastructure scaling beyond discrete chips: AMD now requires Taiwan to manufacture entire AI systems, from silicon through substrates, packaging, server ODMs, and rack integration — reflecting how procurement has evolved into core strategy at scale.
From Components to Complete AI Infrastructure
The significance lies not only in higher order volumes but in systemic integration. As AMD’s business reaches new scale, its needs extend across the full stack — from TSMC’s advanced nodes to final-system integration by Quanta and Foxconn. Su underscored this interdependence in a social media post dated Oct 7, 2026:
“Always wonderful to be back in Taiwan spending time with our customers and partners this week. Thank you for the warm welcome and the tremendous partnership!” — Lisa Su, CEO of AMD
That message signals deepening operational entanglement — where capacity planning now spans three to five years and covers every layer required for AI infrastructure deployment.
Source: tspasemiconductor.substack.com
Compiled from international media by the SCI.AI editorial team.
