US hyperscalers have committed $660B in combined AI capital expenditure for 2026, triggering an 18-day record winning streak on the PHLX Semiconductor Index.1 The scale of this spending — concentrated in American cloud giants — is reshaping global semiconductor supply chains and manufacturing demand from Taiwan to South Korea.
Meta signed a $100B+ chip supply agreement with AMD covering five years.2 Amazon, Alphabet, and Microsoft remain dominant NVIDIA customers, anchoring demand across the world's three largest cloud platforms.2 Together, these contracts represent the largest coordinated AI infrastructure procurement cycle in the industry's history.
Bloomberg Intelligence projects 57% semiconductor revenue growth globally in 2026 — twice the broader market rate.3 Asian chipmakers and foundries, particularly in Taiwan and South Korea, stand to capture significant portions of this capex as fabrication partners to both NVIDIA and AMD.
Meta's AMD deal marks a strategic shift: hyperscalers are diversifying chip supply beyond a single dominant supplier. A five-year contract insulates Meta from spot-market price surges as global AI compute demand accelerates. It also opens revenue opportunities for AMD's international manufacturing and supply partners.
NVIDIA retains dominance in GPU supply for model training. AWS, Google Cloud, and Azure each run large-scale NVIDIA deployments supporting foundation model development and enterprise AI services sold across every major market.
The PHLX's 18-day streak reflects markets pricing in sustained capex execution. Hyperscaler spending guidance has become the primary signal traders worldwide watch for semiconductor revenue visibility. Quarterly capex revisions now move the index within days.
Bloomberg Intelligence's 57% growth forecast doubles the broader market projection.3 That gap reflects AI infrastructure as a discrete demand category — insulated from the consumer electronics cycle that historically drove semiconductor volatility in Asia-Pacific markets.
The $660B commitment positions 2026 as a defining year for global AI infrastructure scale. Capacity built this year will support the next generation of foundation models reaching commercial deployment in 2027 and beyond — with supply chain benefits flowing from Silicon Valley to Hsinchu to Seoul.
