Qualcomm recently held an Investor Day conference where it laid out an aggressive strategy to push into data center market, while working alongside dominant players like NVIDIA and Intel rather than competing with them.

If all goes according to plan, Qualcomm will have diversified its market considerably beyond just smartphones and become a major supplier of AI data center infrastructure and is projecting that its data center business will generate $15 billion in annual revenue within the next few years.

Rather than introducing a single flagship processor, Qualcomm outlined a broad portfolio that includes new server CPUs, AI accelerators, custom silicon capabilities for hyperscale cloud providers, a new high-bandwidth computing architecture and software tools designed to make AI applications portable across multiple hardware platforms.

“The biggest deal for me was all the data center stuff,” said Bob O’Donnell, president and chief analyst with TECHnalysis Research. “It’s a pretty comprehensive story. It’s not a typical ‘here’s our one new chip.'”

Central to Qualcomm’s strategy is assembling all of the pieces needed to compete in modern AI infrastructure. The company introduced:

• New server CPUs aimed at AI workloads, including the Dragonfly C1000.
• Custom silicon design services for major cloud providers, including custom AI ASICs for inference workloads, custom server CPUs based on Qualcomm’s Oryon CPU architecture, and custom SoCs that combine CPUs, AI accelerators, networking, memory subsystems, and interconnects into a single design.
• A new high-bandwidth computing (HBC) architecture designed to reduce dependence on costly high-bandwidth memory.
• Software capabilities gained through its acquisition of Modular that allow developers to write AI applications once and deploy them across multiple hardware architectures.

The software component is particularly significant because it addresses one of the biggest challenges facing new chip architectures: developer adoption. NVIDIA has always known that; it has repeatedly said that the company has many more software developers than hardware developers on staff, and the adoption of its CUDA development language has been key to the company’s dominance in data centers.

“Without software support you’re dead in the water,” O’Donnell said, describing the Modular acquisition as the element that ties Qualcomm’s hardware strategy together.

Qualcomm is not attempting to replicate NVIDIA’s integrated AI systems outright. Instead, executives described a more modular approach in which Qualcomm technology can either operate independently or integrate alongside processors from other vendors.

The company envisions future AI data centers consisting of multiple specialized racks, each optimized for different portions of an AI workload rather than relying on a single monolithic system.

Its HBC architecture is designed to function as an intermediary layer capable of connecting CPUs and GPUs from different vendors while translating workloads between architectures.

In one example described during the briefing, Qualcomm technology could sit between an Intel CPU and an NVIDIA GPU, using Modular’s software translation capabilities to move workloads between the two platforms.

Its strategy received significant validation from both Microsoft and Meta, who are expected to work with Qualcomm on its new Arm-based server CPU efforts. Executives also revealed that two additional unnamed hyperscale cloud providers have signed custom silicon engagements worth more than $1 billion each.

The custom silicon business positions Qualcomm directly against Broadcom and Marvell, companies that have built significant businesses designing custom AI processors for cloud providers.

Outside the data center, Qualcomm also raised its automotive outlook. The company increased its projected fiscal 2029 automotive revenue target from $8 billion to $10 billion, citing growth in its existing design pipeline.

Executives also claimed Qualcomm is on track to become the world’s largest automotive semiconductor supplier, which would be a significant development given entrenched players like NXP, Renesas, Infineon, NVIDIA, and Mobileye.

Qualcomm has long been viewed primarily as a smartphone chip supplier, but that all changes if it can execute on its data center and automotive ambitions. The company expects handset revenue to account for less than 25% of total company revenue by 2029, according to projections shared during the event. In fiscal 2025, smartphone revenue accounted for 63% of total revenue.

And that comes on the heels of losing the Apple business. Apple was a major customer of Qualcomm’s cellular modems for the iPhone, but that all collapsed in a heap in 2017 with a significant lawsuit. The suit was settled in 2019 and Apple began the process of weaning itself off of Qualcomm products. It is estimated that Qualcomm brought in $5 to $6 billion a year just from Apple, and now it is losing that.

O’Donnell warned that Qualcomm’s strategy is considerably more complex than conventional semiconductor product launches and will require further explanation as the products reach market.

The company still faces significant execution challenges in a competitive market, but if it continues to line up significant partners, it could be well-positioned to become a serious contender in the next generation of AI infrastructure rather than simply expanding beyond smartphones.