U.S. Tightening Server CPU Exports to China: Intel, AMD Shrink Shipments, Cloud Shift to Domestic

U.S. restrictions appear extending from GPUs to server CPUs, with Intel and AMD cutting deliveries to China amid heightened supply constraints. Sources show Intel Xeon Scalable and AMD EPYC shipments to China have been sharply curtailed, with delivery lead times rising from about 4 weeks to 16 weeks and prices up 20%–35%. Since December 2025, Intel has implemented quotas on server CPUs in China, reducing single-quarter shipments to China about 30% from 2024. AMD has prioritized shipments to North American cloud providers, sharply reducing available allotments in China, with some x86 parts now available only to U.S. buyers or at a premium. Chinese cloud and telco buyers are adjusting portfolios: Alibaba Cloud reduced x86 share in its server procurements from 70% to 45% in Q1 2026, shifting capacity to Huawei Kunpeng and Phytium processors. China Mobile requires that server procurements in the coming years have at least 50% domestic CPUs, turbocharging Kunpeng and Phytium. Kunpeng server CPU output in January 2026 reached over 800,000 units, about four times 2024 levels, as plants operate three shifts. SeaInfo and other x86 partners are increasing CAPEX by 40% to move into 7nm and below, while Kunpeng prices have risen from about ¥1,800 to ¥2,400 per CPU. The tighter supply is attributed to revised export controls on “high-performance data center processors” in October 2025, not explicitly naming CPUs but prompting Intel and AMD to reduce China-bound deliveries to mitigate compliance risk. Analysts project if constraints continue, China’s server CPU domestic share could rise from 18% in 2024 to 35% in 2026, with an ARM-x86 dual-path architecture emerging. The shift also implies roughly 500,000 additional 12-inch wafers per year for domestic fabs and potential spillovers to equipment and materials suppliers.