Qualcomm wins Meta as a major CPU customer and unveils full AI data center platform (Dragonfly C1000, HBC, AI300), targeting $15 billion data center business by 2029.
Qualcomm has announced a multigenerational CPU deal with Meta, secured two major hyperscale customers, introduced a new AI inference server architecture, and projected over $15 billion in annual data center revenue by fiscal 2029. The announcements at the company's Investor Day event represent its clearest effort yet to position itself as a supplier of AI infrastructure rather than edge devices alone. "We're building a data center platform. It's a comprehensive portfolio of solutions," said Qualcomm President and CEO Cristiano Amon.
The strategy builds on disclosures from Qualcomm's April earnings call, when executives outlined pursuits of hyperscalers, cloud providers, sovereign AI projects, and custom silicon engagements. The company had promised additional customer announcements at Investor Day.
Meta CEO Mark Zuckerberg announced the multigenerational agreement under which Qualcomm will supply CPUs for the company's next-generation server fleet. "Our goal is to deliver personal superintelligence to everyone in the world," Zuckerberg said. "That's why our work with Qualcomm is so critical." Meta declined to discuss deployment timing, processor specifications, or workloads. A Meta spokesperson said the company is "embracing a flexible, portfolio-based approach, combining hardware from a range of partners with our own rapidly advancing MTIA silicon program"—suggesting Qualcomm will become one component of Meta's expanding AI infrastructure rather than replacing the company's in-house silicon efforts.
The Meta deal carries significance beyond immediate revenue, according to Matt Kimball, vice president and principal analyst for data center technologies at Moor Insights & Strategy. "One customer win doesn't change the server CPU market overnight," Kimball said. "But it definitely expands the conversation." He noted that the Meta deal gives Qualcomm revenue to fund future product development and validation from one of the world's largest hyperscale operators, making it easier to pursue additional cloud customers.
Qualcomm executive vice president and general manager of data center Tony Pialis announced that the company has secured two major hyperscale customers. "We have won two major hyperscaler deals that will contribute meaningful revenue to Qualcomm, starting at the end of this year," he said. Chief Financial Officer Akash Palkhiwala stated that the two customers are expected to generate at least $1 billion in revenue within a year. Meta is one of the two hyperscalers; Qualcomm did not identify the second. Microsoft CEO Satya Nadella endorsed Qualcomm's High-Bandwidth Compute architecture during the event but stopped short of announcing a commercial deployment.
Rather than introducing a single server processor, Qualcomm outlined a portfolio spanning CPUs, AI accelerators, networking, custom silicon, and an open software stack strengthened by its acquisition of Modular. Pialis argued that conventional server architectures will not scale to agentic AI. "Traditional infrastructure will not scale to the needs of agentic AI," he said. "The industry needs a paradigm shift." Kimball said Qualcomm's approach distinguishes it from competitors by addressing multiple layers of AI infrastructure rather than a single component. "The CPU/xPU opportunity is huge," he said. "I think there is enough opportunity to support these players – and maybe a couple more."
Kimball also pointed to the Modular acquisition as a potentially important differentiator. If Qualcomm can provide software that efficiently runs AI workloads across multiple silicon architectures, operators could gain greater flexibility as AI infrastructure becomes more heterogeneous.
Central to Qualcomm's strategy is High-Bandwidth Compute, or HBC, which the company says combines SRAM-class performance with HBM-class capacity to reduce memory bottlenecks during AI inference. The company did not disclose performance data or technical implementation details. Kimball said HBC could prove one of Qualcomm's most important technology announcements if it performs as advertised. "If Qualcomm delivers what it's describing, HBC could improve both inference performance and efficiency, particularly in disaggregated AI infrastructure where moving data efficiently is often as important as adding more compute," he said.
Qualcomm's data center business anchors its latest diversification push. Under its revenue forecast, data center would generate more than $15 billion in annual revenue, industrial, networking, and robotics would contribute $8 billion, and handsets would account for roughly one-third of Qualcomm's chip revenue.
Kimball said the data center target is ambitious but achievable if Qualcomm continues adding hyperscale customers. "Hyperscale economics are different than enterprise infrastructure," he said. "A relatively small number of large customer wins can translate into billions of dollars of annual revenue very quickly."