Key Takeaways

  • Qualcomm CEO Cristiano Amon outlined how the Modular acquisition supports the company's multi-vector AI compute strategy for data centers.
  • Amon stated Qualcomm expects to begin generating data center revenue in 2027 as it advances its CPU, accelerator, chiplet, and custom ASIC development for hyperscalers.
  • Industry analysts note that inference-focused hardware is becoming central to AI infrastructure planning, aligning with Qualcomm's roadmap.

Qualcomm is pushing further into the data center market. During a recent conversation on Closing Bell, CEO Cristiano Amon used the company's acquisition of Modular to explain the next phase of Qualcomm's AI strategy. The discussion underscored an ongoing market expansion. Qualcomm has been steadily moving from a primarily mobile focus into a broader compute position spanning PC, automotive, industrial, robotics, and networking environments. The company is now applying that approach to the data center.

Amon reiterated Qualcomm's plan to build a hardware portfolio spanning CPUs, accelerators, and custom chips. Each component supports a different segment of AI infrastructure, aligning with Qualcomm's intention to offer a complete stack rather than a single product. That approach mirrors broader industry behavior. According to IDC, enterprises are increasingly adopting multi-chiplet and heterogeneous architectures to meet AI workload demands without relying on a single class of processor. Amon's comments indicate Qualcomm views this architectural shift as an opening for new competitors.

Qualcomm has stated it expects to begin generating data center revenue in 2027, a timeline closely following the October introduction of its first data center-grade AI chips focused on inference. Inference workloads constitute a growing share of enterprise AI spending. A recent analysis from McKinsey noted that inference workloads already outpace training workloads in many production environments, prompting cloud providers to evaluate power-efficient hardware to support that utilization pattern.

The Modular acquisition directly addresses this software requirement. Modular created tooling and software infrastructure for AI workloads, and folding it into Qualcomm gives Amon an opportunity to pitch a more complete solution to hyperscalers. Software has become a differentiator in compute hardware, and Qualcomm is signaling its intent to compete at the infrastructure layer. Amon also referenced Qualcomm's custom chip capabilities, bolstered by its Alphawave acquisition. Customization remains a growing trend for large AI operators; Deloitte recently reported that hyperscalers are accelerating custom silicon programs to maximize performance per watt and workload-tuned efficiency.

Amon framed these moves as foundational to the next generation of data centers. This refers to a combination of chiplets, novel memory approaches, and AI-tuned accelerators serving both edge and centralized environments. Qualcomm has identified chiplets as a key design framework in its data center plan. Industry watchers point out that chiplet architectures allow engineers to iterate quickly and tailor configurations to specific workloads, potentially lowering the barrier to entry against early leaders in monolithic data center GPU design.

During the discussion, Amon emphasized Qualcomm's priority on inference over training, deliberately placing its early data center designs against a rapidly scaling segment that is less dominated by a small handful of incumbents. Training clusters require high bandwidth and remain overwhelmingly associated with large GPU farms. Inference architectures are more distributed and frequently price-sensitive. Qualcomm is betting its mobile heritage in power efficiency will provide a competitive advantage in inference scenarios. This strategic angle targets cloud providers actively seeking to reduce operational power consumption. Hyperscaler adoption rates will ultimately determine whether this power-efficiency strategy succeeds at scale.

AI adoption continues to pressure existing data center architectures. Many operators have delayed refresh cycles while evaluating which silicon strategies to back. Amon indicated that Qualcomm will deliver alternative architectures as those purchasing decisions are finalized. Qualcomm's established presence in automotive and PC segments provides engineering experience in balancing power, performance, and thermal requirements. Cloud architects suggest those disciplines translate directly to inference clusters requiring high economic efficiency.

Entering the data center market requires extensive testing cycles, deep partnerships, and heavy software investment. Qualcomm's acquisition of Modular supports the software layer, and the Alphawave integration advances its custom chip development. Competing for hyperscaler designs remains a multi-year process. Amon's projection of 2027 revenue indicates the company is structuring its investments for long-term deployments rather than short-term hardware cycles.

The Closing Bell interview outlines Qualcomm's long-term intent to establish itself as a broad compute supplier rather than remaining tied to its mobile reputation. As the industry seeks expanded choices in AI hardware, particularly for inference, Qualcomm's expanding portfolio targets hyperscalers diversifying their silicon strategies. Chiplet designs are lowering entry barriers, and software ecosystems are becoming more modular. Qualcomm intends to leverage Modular's software, its chiplet hardware, and heterogeneous architectures to capture early market share in inference workloads.

The next generation of data centers will rely on heterogeneous, power-efficient infrastructure tuned for AI inference at massive scale. The coming years will demonstrate how much traction Qualcomm can secure across hyperscale deployments as its 2027 commercialization timeline approaches.