Qualcomm takes aim at AI-server demand after signing large hyperscaler arrangements, according to investor-day guidance

Three large hyperscaler-related deals highlighted in a recent market report have drawn attention to a chipmaker outside the usual Nvidia and Intel headline loop. The report says Qualcomm expects its data center chip business to move from “practically nothing” in the prior year to a much larger revenue base over the next several years, supported by custom-silicon efforts with major cloud operators and follow-on product plans.

According to the report, Qualcomm began developing a custom silicon solution with a leading hyperscaler in April and expects to start ramping later this year. In addition, at an investor day the company said it had lined up two hyperscalers under contract for at least $1 billion. The same investor-day framing also included a deal tied to Meta Platforms for Qualcomm’s next-generation CPUs designed for AI agents, with the expectation that selling would begin in the second half of 2028.

The market report further cites Qualcomm’s management outlook for its data center business, saying the company now expects $5 billion in revenue in fiscal 2027, growing to $15 billion by 2029. That guidance is positioned as a response to how AI workloads are evolving, with more buyers looking beyond GPUs toward the broader server CPU and compute stack.

The report’s central theme is that AI compute demand is expanding in multiple directions, not just in graphics processors. It argues that, as “agentic AI” (AI systems that can take actions and execute multi-step tasks) becomes more common, developers are increasingly considering CPUs as part of the workload mix. In that context, the report frames Qualcomm’s custom and next-generation CPU plans as attempts to capture value from data-center spend that goes beyond GPU-only deployments.

While Nvidia remains the dominant name in the public discussion around AI acceleration, the report explicitly steers readers away from treating Nvidia or Intel as the only beneficiaries. It instead describes Qualcomm as trying to build a new revenue line in data center silicon, supported by contracts with large cloud customers and a longer-dated roadmap for agent-focused CPUs.

Qualcomm’s opportunity, as described in the reporting, sits at the intersection of two trends: custom chips for hyperscalers and the need for specialized server CPUs tailored to the performance and power demands of inference and AI services. Custom silicon can matter to hyperscalers because it can be optimized for specific software stacks and performance targets, potentially improving economics compared with using more general-purpose components at scale.

Still, key details are not provided in the market report itself. It does not break out deal sizes by customer, does not describe the technical specifications of the custom silicon or agent-focused CPUs, and does not clarify whether the “three massive deals” referenced are all new contracts or a combination of arrangements described across different company updates. Readers looking for more certainty would need Qualcomm’s primary disclosures tied to the April development and the investor day where the contracts and revenue targets were discussed.

Investors and industry watchers will likely focus next on Qualcomm’s disclosed milestones for ramp timing, confirmation of how much of the $5 billion fiscal 2027 target is tied to which hyperscalers, and any additional product announcements around the second-half-2028 commercialization timeline for the AI-agent CPU generation. Monitoring data center revenue disclosures and management commentary during upcoming earnings and updates will be especially important given the multi-year nature of the forecast.