TL;DR
Qualcomm unveiled the Dragonfly C1000 server CPU at its June 2026 Investor Day, locked in Meta as a multi-generation customer, and nearly doubled its non-handset revenue target to $40 billion by fiscal 2029, signaling a serious pivot from smartphone chips to AI data center infrastructure.
What happened
- Dragonfly C1000 is a server CPU built for agentic AI workloads, using a multi-chiplet design with more than 250 Oryon cores running above 5 GHz.
- Qualcomm claims more than double the performance per watt of competing server CPUs, a critical edge as power constraints increasingly cap data center expansion.
- Meta Platforms signed a multi-generation agreement to deploy the C1000 in its data center servers, with volume production slated for the second half of 2028.
- Microsoft is expected to deploy Qualcomm's High Bandwidth Compute (HBC) chips in Azure, and additional unnamed hyperscalers may contribute custom-chip revenue before end of 2026.
- Qualcomm raised its non-handset revenue target to $40 billion for fiscal 2029, up from a prior target of $22 billion, and set a $15 billion-plus data center revenue goal for the same year.
Why it matters
- Shares jumped more than 12% in after-hours trading and Morgan Stanley upgraded QCOM to Equal-Weight, raising its price target to $231 from $146, reflecting a market that had underpriced the data center opportunity.
- The $200 billion CPU market Qualcomm is targeting by fiscal 2029 or 2030 dwarfs its current smartphone base, and the company projects handsets will fall to only one-third of total revenue by then.
- Most chip companies announce server ambitions without customers. A signed Meta deal plus expected Microsoft deployment gives Qualcomm two of the largest compute buyers as early validators before the product even ships at scale.
- Qualcomm is also acquiring Modular Inc. to build an open, cross-architecture AI software stack for data center and edge, closing in H2 2026, which addresses the software gap that has historically undercut ARM-based server challengers.
- The performance-per-watt argument is structurally timed well: hyperscalers are increasingly power-constrained, making energy efficiency a purchasing criterion that favors Qualcomm's pitch over raw x86 throughput.
What to watch next
- Fiscal 2027 data center revenue of $5 billion is the first hard checkpoint for whether the $15 billion-plus fiscal 2029 target is on track, watch quarterly earnings for early signals.
- Volume production in H2 2028 is still two years out, leaving room for Intel, AMD, or Nvidia (via its Vera chip) to respond with competing efficiency claims or to lock in alternative hyperscaler commitments.
- Unnamed hyperscaler customers contributing custom-chip revenue before end of 2026 would confirm that Meta and Microsoft are not one-off wins but the start of a broader platform shift.
Originally published on Present of AI, a daily source-linked AI news timeline. Read the full timeline or browse the open dataset.