Qualcomm enters server CPUs amid AI demand
- Qualcomm said on April 29 it had entered the data center market, with a leading hyperscaler custom-silicon engagement scheduled for initial shipments later in 2026. - Nvidia said on May 18, 2025 that Qualcomm planned custom CPUs linked to Nvidia GPUs through NVLink Fusion for semi-custom AI infrastructure. - Qualcomm said it will give investors a data-center update at its June 24 Investor Day, including growth initiatives in Data Center and Physical AI.
Qualcomm has already moved beyond rumor in server processors. The company said on April 29 that it had entered the data center market and that a “leading hyperscaler custom silicon engagement” was on track for initial shipments later in 2026. Chief Executive Cristiano Amon tied that push to the rise of AI agents and said Qualcomm would provide a fuller update at its June 24 Investor Day. Nvidia had signaled Qualcomm’s server-CPU plans earlier. On May 18, 2025, Nvidia said Qualcomm and Fujitsu each planned to build custom CPUs that would be coupled with Nvidia GPUs using NVLink Fusion and Spectrum-X networking in semi-custom AI systems. Qualcomm later said at Computex 2025 that its “advanced custom CPU technology” with Nvidia’s AI platform would target data center infrastructure. (qualcomm.com) ### What did Qualcomm actually say? Cristiano Amon said in Qualcomm’s fiscal second-quarter release that the company was “equally excited” by its entry into the data center and that shipments for the hyperscaler engagement were expected later this calendar year. The same release said Qualcomm would discuss Data Center and Physical AI at Investor Day on June 24. (investor.nvidia.com) Qualcomm did not name the hyperscaler in that release. The company also did not disclose pricing, unit volumes or a launch date for a standard merchant server CPU product in the materials reviewed. ### Where does the server CPU angle come from? Nvidia’s May 2025 NVLink Fusion announcement put Qualcomm in the group of companies planning custom CPUs connected to Nvidia GPUs. (qualcomm.com) That matters because NVLink Fusion is designed for hyperscalers and custom chip designers that want to integrate their own CPUs or accelerators into Nvidia-based rack systems. Qualcomm’s own Computex commentary pointed in the same direction. In a company blog post after Amon’s keynote, Qualcomm said it saw room “not only at the edge, but also in the data center,” and said its custom CPU technology with Nvidia’s full-stack AI platform would bring computing to data center infrastructure. ### Is there evidence that AI demand is lifting CPUs, not just GPUs? (investor.nvidia.com) Intel said on April 23 that demand was running ahead of supply “especially for Xeon server CPUs,” and CEO Lip-Bu Tan said the company expected sustained momentum this year and next. Intel’s prepared remarks did not, in the material reviewed, give a 1:1 CPU-to-GPU ratio, but they did say Xeon demand was outpacing supply. (qualcomm.com) AMD said on May 5 that first-quarter 2026 data-center revenue rose 57% year over year. Third-party transcript summaries of AMD’s call said management now saw server CPU total addressable market growth above 35% annually to more than $120 billion by 2030, driven by agentic AI demand, but that specific formulation was not confirmed in AMD’s primary press release reviewed here. (download.intel.com) ### Can the claims about lead times and price increases be verified? The May 14 X post cited in the prompt said server CPU lead times were eight to 12 weeks and that Intel and AMD had raised prices. Reuters-style verification requires primary sourcing for those claims, and that was not available in the official materials reviewed. Intel did say on April 23 that demand continued to run ahead of supply for all its businesses, especially Xeon server CPUs. (ir.amd.com) That supports the existence of tight supply, but not the specific eight-to-12-week lead-time claim or any specific recent price increase by Intel or AMD. ### Why does Qualcomm’s move matter now? Qualcomm is re-entering a market it left years ago, this time with custom CPU technology tied to AI infrastructure rather than a broad stand-alone server push. (qualcomm.com) The company’s public comments so far center on a named hyperscaler engagement, Nvidia interoperability and later shipments in 2026. June 24 is the next date to watch. (download.intel.com) Qualcomm said it would update investors that day on growth initiatives including Data Center and Physical AI, which is the company’s next scheduled forum for more detail on customers, products or timing. (qualcomm.com)