Qualcomm announced a comprehensive data center AI strategy on June 24, projecting $15 billion in data center chip revenue by fiscal 2029 (ends September 2029), with $5 billion arriving in fiscal 2027. The company secured Meta and Microsoft as customers for its new Dragonfly data center chips, plus two unnamed hyperscalers for custom ASICs with revenue starting before year-end 2026. Qualcomm also raised its broader non-handset revenue target to $40 billion by FY29, roughly double prior guidance of $22 billion.
Qualcomm stock surged 12% in after-hours trading following the announcement. CFO Akash Palkhiwala stated the $5 billion fiscal 2027 data center total includes $1 billion from custom-chip customers. The company is targeting more than $18 non-GAAP EPS by fiscal 2029. Qualcomm also acquired AI software company Modular for approximately $3.92 billion in an all-stock transaction to compete with Nvidia's CUDA ecosystem and help customers run models across different chip architectures without rewriting code.
The shift underscores Qualcomm's pivot away from smartphone dominance—where memory shortages constrain growth—into AI data center, automotive (now >$5B annualized revenue run-rate), and robotics. Management expects handsets to represent roughly one-third of total chip revenue by FY29, down from historical dominance. Competition is fierce: Nvidia dominates AI GPUs, Broadcom leads custom chips, and hyperscalers (Google, Amazon) design their own processors.
For architects, Qualcomm's Dragonfly CPUs and inference accelerators—targeting production shipments in late 2026—represent a second credible alternative to Nvidia for non-training workloads. The $15B target signals management conviction that the data center TAM can absorb multiple players, though hyperscalers' own silicon and Nvidia's moat remain formidable. Timeline: Dragonfly C1000 chip for Meta ships in 2028, underscoring a multi-year ramp.