Qualcomm unveils Dragonfly C1000 data-center CPU; Meta commits to 2028 production volumes

Qualcomm on Wednesday announced Dragonfly C1000, a CPU designed for agentic AI workloads in data centers, with Meta committing to use the chip when production begins in 2028. CEO Cristiano Amon presented the roadmap at Qualcomm's investor day, signaling the chipmaker's aggressive pivot into the hyperscaler market beyond smartphones. The company secured custom-silicon deals with nearly every major cloud provider through existing edge and smartphone relationships.

Dragonfly C1000 is built around Qualcomm's expertise in low-power, battery-efficient processors—a design advantage the company believes translates to data-center economics where electrical power is increasingly the limiting factor. Qualcomm also announced the acquisition of Modular, a startup that built AI-enabling software for multiple chip architectures (equivalent to NVIDIA's CUDA ecosystem). The company secured two custom-silicon deals with hyperscalers and is pursuing diversity into cars, robots, and now data centers as smartphone shipments have plateaued.

Qualcomm's pitch centers on engineering scale and operational execution rather than late entry. CFO Akash Palkhiwala noted the company already has relationships with every major hyperscaler through edge products, positioning Dragonfly as an extension rather than a new relationship. Shares were down on Wednesday, likely reflecting long-standing skepticism about Qualcomm's execution in new markets.

For infrastructure architects, Dragonfly + Modular matters because Qualcomm is explicitly building CUDA parity across CPU architectures. If Modular's software layer works as intended, it reduces switching costs from NVIDIA GPUs to Qualcomm CPUs for agentic workloads. Meta's 2028 commitment suggests real traction, though the delayed ramp (two years out) gives NVIDIA and AMD time to respond with competing CPU offerings.