As AI data centre growth intensifies, Intel and AMD have locked Chinese server operators into multi-year CPU supply agreements, reflecting escalating prices and tightening availability.
- Multi-year CPU purchase commitments signed by Intel and AMD in China
- Server CPU prices in China up over 40% year-to-date amid supply tightness
- Demand driven by AI data centre expansion and export restrictions on Western hardware
What happened
Intel and AMD have entered into extended supply agreements with Chinese data centre customers, committing to multi-year purchase volumes for their server CPUs. These contracts prioritize guaranteed volumes over fixed pricing, allowing chipmakers to adjust prices as supply pressures continue.
This shift comes amid soaring demand for general-purpose processors in Chinese data centres, where access to advanced Western AI accelerators is limited due to U.S. export controls. Consequently, operators have intensified their reliance on CPUs from Intel’s Xeon and AMD’s EPYC lines to support AI workloads and overall infrastructure.
Why it matters
The server CPU market in China is undergoing rapid change as prolonged supply shortages push prices sharply higher—some models have surged by more than 40% since the start of the year. This is a notable departure from the historical stability of the segment, traditionally overshadowed by GPU demand.
These developments indicate that the AI boom’s hardware demand now extends beyond accelerators to encompass all supporting components, including processors, memory, and networking equipment. For Intel and AMD, these volume commitments represent both a strategic win and a challenge to meet escalating demand and manage fluctuating prices.
What to watch next
Monitoring the duration and pricing mechanisms of these multi-year contracts will be critical in assessing how Intel and AMD manage supply shortfalls and pricing volatility amid rising demand from China's AI infrastructure expansion.
Further industry trends to watch include how local Chinese chip designers adapt to these constraints, the potential development of custom ASICs, and whether supply chain pressures spread to other server components beyond CPUs and memory.