Qualcomm has spent four decades as the chipmaker inside the world's smartphones. On Wednesday in New York, it tried to convince Wall Street it can also be the chipmaker inside the world's data centers — and it brought Mark Zuckerberg along to make the case.
At its 2026 Investor Day at the Ziegfeld Ballroom in Manhattan, Qualcomm unveiled a sweeping data-center roadmap anchored by a new server processor, the Qualcomm Dragonfly C1000 CPU, and announced that Meta had signed on as a strategic, multi-generation customer. It is the first named buyer for Qualcomm's data-center push, and the endorsement of one of the largest infrastructure spenders on the planet instantly reframed a years-long gamble as a credible challenge to Nvidia, AMD and Arm in the most lucrative corner of the chip market.
A device partner becomes a data-center customer
The agreement commits Meta to deploy the Dragonfly C1000 — and its successor generations — across its next-generation server fleet. Qualcomm said the chip will enter production in the second half of 2028, a timeline that makes clear this is a long-horizon bet rather than a near-term revenue event.
"We designed our data center CPU to deliver leading performance per core and a breakthrough in power efficiency for large scale data center deployments, and this multi-generation agreement with Meta is a significant validation of that approach," said Cristiano Amon, President and CEO of Qualcomm Incorporated. "We're thrilled to build on our partnership with Meta, expanding from devices to data center. And this is just the beginning."
Zuckerberg, who appeared at the event, framed the deal as one piece of Meta's escalating infrastructure build-out. "We're excited to continue partnering with Qualcomm Technologies as they design the next generation of CPUs for Meta," said the Meta founder and CEO. "Along with our other compute investments, we're quickly building the infrastructure we need to deliver personal superintelligence to everyone in the world."
The Dragonfly C1000 is an Arm-based server CPU built around Qualcomm's custom Oryon cores running past 5GHz, with a chiplet design that packs more than 250 cores per chip. Qualcomm is pitching it on performance per watt — the metric that increasingly governs total cost of ownership as operators run up against power and cooling limits at hyperscale.
The full Dragonfly stack
The CPU was the centerpiece, but not the whole pitch. Qualcomm also detailed a multi-generation line of AI inference accelerators: the AI200, slated to begin customer shipments later this year; the AI250, due in 2027; and a newly added AI300, which is expected to sample in 2028. The company tied the lineup to a new memory architecture it calls High Bandwidth Compute (HBC), claiming the AI250 delivers an 18x effective-bandwidth improvement over the AI200, with the AI300 targeting 54x.
Notably, these accelerators are aimed squarely at inference — running trained models at scale — rather than the training workloads that Nvidia's GPUs dominate. That is a deliberate flanking maneuver. As AI shifts from one-time training runs toward always-on "agentic" applications, the volume of inference compute is exploding, and Amon argued that environment plays to Qualcomm's roots in low-power silicon. As AI agents drive demand for inference, he said, "infrastructure has to deliver much higher performance at lower power and cost."
Qualcomm rounded out the day with a roughly $3.9 billion all-stock acquisition of AI software firm Modular and an expanded partnership with Hugging Face to bring open-source models onto Dragonfly hardware — a tacit acknowledgment that selling silicon to hyperscalers requires a software ecosystem to match Nvidia's CUDA moat.
Big targets, distant payoff
The financial framing was aggressive. Qualcomm raised its fiscal 2029 non-handset revenue target to $40 billion, roughly double its prior goal, and set a data-center-specific target of more than $15 billion by fiscal 2029. Executives described "multiple inflection points over the next three to five years" as the company diversifies away from a smartphone market that has long defined — and limited — its growth.
Investors were enthusiastic, then ambivalent. QCOM shares spiked as much as 11% to around $219 on Thursday before retreating, and by later sessions traded near $193, with the stock weighing the near-term cost of the Modular deal against data-center revenue that will not meaningfully arrive until late 2028. A wave of analyst price-target hikes followed the announcements.
For Meta, the deal is less about Qualcomm and more about a deliberate multi-vendor strategy. The company has been designing its own MTIA accelerators, buys heavily from Nvidia, and licenses Arm architecture — and reports surfacing alongside Qualcomm's event suggested Microsoft is also in the mix as a customer. Adding Qualcomm CPUs gives Meta another lever against single-supplier pricing power and supply constraints, the same calculus driving every hyperscaler to cultivate alternatives to Nvidia.
What to watch
The hard questions start now. A 2028 production date means roughly two years before the first dollar of Meta revenue lands, and the gap between an investor-day roadmap and silicon running in production racks is where ambitious chip programs often stumble. Qualcomm must prove the C1000 hits its performance-per-watt claims, that the AI200 ships on schedule this year, and that the Modular and Hugging Face bets can blunt Nvidia's software advantage. Watch, too, whether Meta's commitment translates into disclosed order volumes — and whether any second or third hyperscaler signs publicly. Qualcomm has finally planted a flag in the data center. The next two years will decide whether it can hold the ground.
"We're thrilled to build on our partnership with Meta, expanding from devices to data center. And this is just the beginning."- Cristiano Amon, President and CEO, Qualcomm Incorporated