Qualcomm Lays Out New Data Center Roadmap For AI; Meta Buys It

In a move that upends the AI chip landscape, Qualcomm has secured Meta as a cornerstone customer for its newly outlined data center roadmap. At its Investor Day, the mobile chip giant announced a multi-year agreement to deploy its upcoming custom AI accelerators in Meta’s data centers, marking Qualcomm’s most aggressive—and credible—push into the server market in years.

Qualcomm, long synonymous with smartphone processors, has tried and failed to break into data centers before. Its Centriq ARM-based server chips were shuttered in 2018 after struggling to gain traction. But the artificial intelligence wave has rewritten the rules. The explosion of large language models and generative AI workloads has created a voracious appetite for inference chips—specialized processors that run already-trained models—and that is the segment Qualcomm is targeting with its new roadmap.

The company revealed the broad strokes of its data center AI strategy during the investor event, but the headline was the Meta deal. Terms were not disclosed, but the deal is described as “multi-year” and involves Meta deploying Qualcomm’s technologies across its global infrastructure. The chips are designed specifically for AI inference, handling tasks like serving up chatbot responses, powering recommendation algorithms, and running image generation models. Meta, which operates one of the world’s largest fleets of servers, has been investing heavily in custom AI silicon—including its own MTIA chips—but continues to rely on external suppliers to meet demand.

For Qualcomm, this agreement is a validation of its pivot. The company’s AI data center roadmap is built on decades of expertise in low-power, high-efficiency computing from its mobile business. Qualcomm claims its upcoming accelerators will dramatically reduce power consumption per inference compared to competing GPUs, a critical selling point as data center energy costs balloon. The first chips are expected to sample later this year, with volume deployments starting in 2027.

Industry observers see this as a direct challenge to NVIDIA, which dominates AI training and inference with its GPUs. However, the inference market is still fragmented, and Qualcomm’s warm reception from Meta—a company known for building everything from servers to networking gear in-house—suggests the chips offer something unique. Still, the road ahead is fraught. Qualcomm must prove it can deliver at scale, secure additional anchor customers, and navigate a chip market plagued by supply-chain bottlenecks and geopolitical tensions.

“This is Qualcomm’s best shot at becoming a major player in the data center,” said one semiconductor analyst who tracks the sector. “Meta is a demanding partner. If they’ve signed on, it’s because Qualcomm convinced them the technology is real.”

Looking ahead, all eyes will be on Qualcomm’s ability to broaden its customer base beyond Meta. The company is expected to target hyperscalers like Microsoft, Google, and Amazon—all of which are developing their own AI silicon but still purchase off-the-shelf chips for certain workloads. The 2027 deployment timeline gives Qualcomm room to refine its architecture and line up additional design wins. For now, the belief that Qualcomm can finally compete in the data center has been given a powerful boost by Meta’s endorsement.