Key Takeaways

  • Qualcomm outlined a data center revenue goal of more than $15 billion by fiscal 2029.
  • The company is positioning its Dragonfly CPUs and AI inference accelerators to compete with Nvidia and AMD.
  • Growing demand for power-efficient AI infrastructure is creating openings for alternative architectures.

Qualcomm has set a data center revenue target of more than $15 billion by fiscal 2029, a move that signals how aggressively it plans to expand beyond its mobile roots. At the moment, this business is small, bordering on nonexistent, but the strategic intent is unmistakable. The announcement lands at a time when the competitive dynamics around AI infrastructure are shifting fast.

The broader market context helps show why Qualcomm is pushing into this sector. The global AI semiconductor market for data centers is projected to exceed $200 billion by 2030, according to McKinsey. That figure reflects combined demand for both training and inference silicon, though inference is expected to grow faster as models spread into production systems. Qualcomm is leaning into exactly that part of the stack. Instead of challenging Nvidia in training, it is betting that data center operators will care more about power efficiency and architectural diversity as their AI fleets scale.

Hyperscale power consumption keeps climbing. The International Energy Agency has estimated that global hyperscale data centers could reach more than 35 gigawatts of energy consumption by 2030. That expectation is one of the primary reasons vendors with low-power designs are gaining attention. Qualcomm frames its AI200, AI250, and AI300 accelerators as efficient inference engines designed to handle large language model serving without the thermal footprint associated with top-tier training GPUs.

The silicon is only part of the story. Qualcomm is also building a software stack intended to compete with Nvidia’s CUDA and NCCL frameworks. The company’s acquisition of Modular was aimed at creating a CUDA alternative that developers can adopt without giving up performance or established workflows. Building a competing software ecosystem will take time, but the investment reflects a long-term infrastructure strategy.

Meta and Microsoft have been cited as early ecosystem partners in Qualcomm’s data center strategy. Their participation adds weight, even if details about deployments remain limited. Hyperscale operators are already deploying heterogeneous compute architectures, often blending custom CPUs, GPU fleets, and domain-specific accelerators. This trend aligns with research from Gartner, which indicates Arm-based and custom CPUs could capture more than 15% of cloud data center processor shipments by 2026. Qualcomm’s Dragonfly CPU family targets this shift, combining general compute and AI inference silicon to appeal to operators seeking alternatives.

Data center AI systems increasingly rely on interoperability standards built around PCIe, NVLink-class interconnects, and emerging open accelerator interfaces. These standards matter because proprietary interconnects are one of the primary reasons Nvidia continues to dominate. Nvidia controls roughly 80% to 90% of the training market, according to Omdia, and its ecosystem advantages are heavily tied to software and networking integration. Qualcomm, by contrast, is pitching an approach aligned with broadly adopted standards and open interfaces, aiming to attract buyers requiring architectural flexibility.

Questions remain about how cloud providers will balance high-performance training clusters with massive fleets of inference servers. Will they prioritize a single vendor for operational simplicity, or diversify architectures to reduce risk and cost? Qualcomm is betting on diversification. The rising share of inference workloads, highlighted in reports from IDC, suggests that many enterprises will look for optimized silicon that scales more efficiently than large training GPUs.

AI workloads are turning into a material portion of capital expenditure budgets, and procurement teams are exploring alternatives to reduce supply chain exposure. Nvidia has historically seen long lead times for its top accelerators, and AMD’s Instinct MI series is becoming a common alternative. Qualcomm hopes to turn that buyer readiness into a foothold. Even a modest percentage of the expanding AI server market could generate billions in revenue, making the company's $15 billion target ambitious but mathematically viable.

Qualcomm still faces challenges in this expansion. Building a competitive developer ecosystem takes time, and convincing enterprises to adopt a new compute architecture requires sustained effort. Silicon roadmaps need to land on schedule, and early partners like Meta and Microsoft will expect predictable performance. However, Qualcomm’s strategy aligns with several long-term data center trends: heterogeneous compute, rising inference workloads, and the urgent need for power-efficient architectures that fit within tightening energy constraints.

By setting a clear revenue target and sharpening its focus, Qualcomm is signaling that its AI data center push is a permanent strategic shift. It is an attempt to secure a slice of one of the fastest-growing segments in enterprise infrastructure. The next few years will determine how much market share it can capture, and whether a mobile-era leader can establish a permanent footprint in the data center.