China: Race to Secure Server CPUs from Intel and AMD at Any Cost
After an initial phase of substantial investments in accelerators, CPUs are also enjoying an unprecedented demand from the data center sector for artificial intelligence. In China, it appears that Intel and AMD are signing large agreements that could ensure hefty profits.
According to reports from Reuters, citing sources familiar with the matter, Intel and AMD have signed longer-term supply agreements with Chinese clients in the server sector compared to previous deals, amid rising prices for CPUs intended for data centers. These contracts are said to establish purchase volumes but not prices, a structure that exposes buyers to the risk of further price increases if the scarcity of components persists.
The Chinese market represents a special case, also due to export restrictions imposed by Washington over the past two years. With access to the most advanced Western accelerators heavily limited, Chinese operators have decisively shifted towards components still freely available, including server CPUs from Intel and AMD.
This scarcity has already generated significant price phenomena: servers based on NVIDIA B300, for example, have reportedly been traded in China for figures close to one million dollars, roughly double the average price in the United States. The same restrictions have pushed Chinese chip designers towards custom ASIC solutions, which, however, do not eliminate the need for a conventional processor to manage the system.
In the past, Chinese customers typically committed to about a year of supply; according to sources cited by Reuters, both manufacturers are now discussing with some buyers commitments of two years or more. Neither Intel nor AMD responded to requests for comment, and the identities of the involved customers have not been disclosed.
The most significant aspects of the agreements remain confidential: actual volumes, pricing formulas, and the total number of buyers involved. What is evident is the trend direction, which would involve both Intel's Xeon line and AMD's EPYC products.
The described price increases are significant: some server CPUs in China have reportedly seen monthly increases exceeding 10%, with some lines rising over 40% since the beginning of the year, a trend more typically associated with the memory market rather than processors.
The underlying cause would be the same one reshaping the entire hardware sector: the expansion of data centers dedicated to artificial intelligence has driven demand well beyond NVIDIA accelerators, involving also memory, networking devices, and the CPUs that orchestrate the operation of racks.
Delivery times for some Intel Xeons have reportedly stretched to about six months. It should be clarified that this phenomenon is not limited to China or just CPUs; thus, processors represent the last link in a supply chain already under pressure, and the multi-year contracts signed in China can be seen as an attempt to preempt further bottlenecks.
For Intel and AMD, this is, in any case, a positive issue. In April, Intel CEO Lip-Bu Tan told analysts that demand continued to exceed supply, a statement that has proven to be even cautious over time. AMD, for its part, expects the server CPU market to exceed $120 billion by 2030.
The structure of the agreements, based on volumes rather than prices, speaks volumes on its own: buyers would be willing to guarantee two years of orders without knowing the final price in advance, betting that availability, rather than cost, is the element that truly needs protection.