AMD’s Trans-Pacific Capacity Race: $7B Arizona Bet Meets Taiwan’s 2nm Surge

Advanced Micro Devices is pursuing a dual-geography strategy to secure the packaging and fabrication capacity needed to feed its surging data-center business, with simultaneous moves in Arizona and Taiwan that underscore the scale of the AI infrastructure buildout. On one side, chip-packaging specialist Amkor Technology is investing up to $7 billion in a new Arizona campus dedicated to advanced packaging for AMD. On the other, AMD chief executive Lisa Su has been in Taipei pressing partners such as ASE, Wistron and Unimicron for drastic capacity increases through 2029. The twin efforts reflect a supply chain under extreme strain.

The urgency is rooted in AMD’s accelerating revenue trajectory. The company posted first-quarter 2026 revenue of $10.3 billion, a 38 percent year-over-year jump, with the data-center segment alone climbing 57 percent to $5.8 billion on the back of EPYC processors and rising Instinct GPU shipments. Management has guided for roughly $11.2 billion in the current quarter. That growth pace is stretching every link in the production chain — from substrates and third-party manufacturing to component availability, all of which AMD flagged as risk factors in its quarterly communication.

Amkor’s Arizona project aims to close one of the most acute bottlenecks: the packaging of multi-die chips for AI workloads. The company has secured 67 additional acres adjacent to its existing 104-acre site in the state, where it will build a campus in two phases. The first fabrication plant is scheduled for completion by mid-2027, with production beginning in early 2028. Amkor itself expects revenue of $8.5 billion to $9.5 billion in 2028 and $11 billion by 2030, but any direct financial impact on AMD remains years away. Still, the symbolic value is immediate: the deal strengthens AMD’s onshoring narrative and gives investors a tangible asset in the US supply chain.

Should investors sell immediately? Or is it worth buying AMD?

Across the Pacific, AMD is leaning even harder on Taiwan’s semiconductor ecosystem. The next-generation EPYC processor, codenamed “Venice,” will be manufactured by TSMC on the industry’s first high-performance 2-nanometer node. Meanwhile, the company is expanding its advanced packaging capabilities in China: Su unveiled the second phase of the TF-AMD facility in Suzhou this week, a move designed to secure capacity for scaling AI infrastructure in the second half of 2026. From 2027, AMD plans to significantly increase server-market delivery volumes.

The technical focus is shifting. AMD is no longer targeting pure AI training alone; inferencing and agent-based AI are taking center stage as hyperscale customers pay closer attention to total system cost. That trend is driving demand for more energy-efficient processors, a sweet spot for the upcoming Venice lineup.

The market has responded with enthusiasm. AMD shares hit a new 52-week high of €394.60 on the back of the Amkor news, representing a gain of roughly 106 percent year to date. The stock now trades at a price-to-earnings ratio of approximately 147, reflecting lofty expectations. Evercore ISI analysts reacted by lifting their price target to $579, a level that implies further upside even after the share price doubling since January. Whether the combined capacity expansion in Arizona and Taiwan can keep pace with AI demand will become clear when Amkor’s campus fires up in 2028 — but for now, AMD is placing its bets on both sides of the Pacific.