Dylan Patel, chief analyst at SemiAnalysis, a leading industry firm, stated, “In the past six months, the entire cloud CPU market has been completely overwhelmed. There’s no capacity anywhere.” Amazon’s deployment of CPU servers this year has tripled year-on-year. Microsoft, on the other hand, sold its own lab CPUs to Anthropic and OpenAI, even experiencing instability issues with GitHub.
TrendForce data confirms this: currently, the CPU to GPU ratio in AI data centers is approximately 1:4 to 1:8, but with the proliferation of AI agents, this ratio will shrink to 1:1 to 1:2. Evercore ISI analyst Mark Lipacis even predicts that in the future, we might see one GPU paired with two to ten CPUs, with CPU growth significantly outpacing GPU growth.
The CPU shortage isn’t just a concept—the supply chain is already seeing real price increases. According to data from StockWe.com, AMD and Intel’s 2026 server CPU production capacity is essentially sold out, and both plan to raise prices by 10% to 15%. AMD is even more aggressive, preparing to raise prices by a cumulative 16% to 17% in two phases during Q2 and Q3.