IMPLICATOR
Qualcomm told investors Wednesday that it expects more than $15 billion in data center revenue by fiscal 2029. In a same-day announcement, Meta agreed to use Qualcomm's Dragonfly C1000 CPUs in servers scheduled for production in the second half of 2028. The target sits inside a broader forecast that lifts Qualcomm's fiscal 2029 non-handset chip revenue to $40 billion, nearly twice its prior $22 billion goal, and would leave handsets at about one-third of QCT revenue, the company's chip business. Shares rose more than 12% in after-hours trading after the New York investor day, Reuters reported.
The June 24 update puts Meta on the CPU side of Qualcomm's data center roadmap. Last October's AI200 and AI250 rollout centered on Humain, the Saudi buyer planning 200 megawatts of accelerator capacity from 2026. The Meta agreement adds the first public CPU customer for C1000 server processors scheduled for 2028.
Key Takeaways
- Qualcomm now targets more than $15 billion in data center revenue by fiscal 2029.
- Meta will use Dragonfly C1000 CPUs scheduled for production in the second half of 2028.
- Reuters valued Qualcomm's all-stock Modular deal at $3.92 billion based on 19.2 million shares.
- AI250 HBC sampling is expected in mid-2027, with AI300 sampling planned for 2028.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Meta puts a name on the CPU customer list
Qualcomm listed more than 250 custom Oryon cores for the Dragonfly C1000, with clock speeds above 5 GHz. The chiplet CPU also supports PCIe Gen 7, CXL connectivity and optional High Bandwidth Compute attach. Qualcomm says its published specifications show more than twice the performance per watt of comparable server CPUs.
The releases did not include an independent C1000 benchmark.
Meta's release says the first C1000 systems are planned for production starting in the second half of 2028, along with future capacity expansions. Zuckerberg said Meta was "quickly building the infrastructure" for personal superintelligence and would use Qualcomm's next-generation CPUs as part of that effort. The companies called the supply deal multi-year and multi-generation; the releases did not include volume, price or workload details.
Tony Pialis, Qualcomm's data center chief, told investors the company also has two unnamed hyperscaler customers for custom chips, with revenue expected before the end of this calendar year. CFO Akash Palkhiwala told Reuters the fiscal 2027 target is $5 billion in data center revenue. About $1 billion of that would come from new custom-chip customers.
Modular is the software answer
Qualcomm is buying Modular Inc. Chris Lattner and Tim Davis co-founded the startup in 2022, and WIRED reported that both are expected to join Qualcomm when the deal closes. Qualcomm's release did not name a price. Reuters calculated a $3.92 billion value from Qualcomm's last closing share price and the up to 19.2 million shares offered to Modular equity holders. The closing window is the second half of 2026, subject to approvals.
Get Implicator.ai in your inbox
Strategic AI news from San Francisco. No hype, no "AI will change everything" throat clearing. Just what moved, who won, and why it matters. Daily at 6am PST.
No spam. Unsubscribe anytime.
Qualcomm said Modular's stack is designed to run AI models across CPU, GPU, NPU and custom ASIC architectures without code rewrites. The Next Web described the product set as the Mojo programming language and MAX inference engine. Amon framed the deal as support for "developer-friendly, horizontal platforms" across different compute environments. Reuters described CUDA as the Nvidia platform that ties developers to its chips.
WIRED put Modular's team at roughly 150 people. The magazine said Lattner created LLVM and Apple's Swift programming language before later work at Google, and that Davis co-created TensorFlow Lite. Qualcomm had earlier bought Ventana Micro Systems, a RISC-V server CPU startup.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
The 2029 target now has checkpoints
Qualcomm put CPUs, inference accelerators, custom silicon and connectivity on the Investor Day roadmap. The AI250 accelerator with first-generation HBC is expected to sample in mid-2027, and Qualcomm said the card is designed for 133 TB/s of effective memory bandwidth. The AI300 with second-generation HBC is expected to sample in 2028, with a claimed 54-fold bandwidth increase over AI200.
Reuters' competitive list for Qualcomm included Nvidia, Cerebras, Amazon's Graviton CPU and Google's Axion CPU. The same report cited Broadcom and Marvell in custom ASICs, one of the three chip categories Qualcomm says it is pursuing with hyperscalers. Qualcomm's release framed Dragonfly around inference workloads, performance per watt and token throughput at lower total cost of ownership.
Qualcomm gave investors several dated checkpoints. Modular is expected to close in the second half of 2026, and Pialis told investors custom-chip revenue should start before calendar year-end, Reuters reported. HBC sampling is due in mid-2027, AI300 sampling in 2028 and Meta's C1000 production in the second half of 2028. Qualcomm attached the $15 billion data center revenue target to fiscal 2029, after those product and customer milestones are scheduled to have started.
Frequently Asked Questions
What did Qualcomm announce at its investor day?
Qualcomm set a fiscal 2029 data center revenue target above $15 billion, raised its non-handset chip revenue target to $40 billion, named Meta as a Dragonfly C1000 CPU customer, and announced a deal to buy Modular.
When will Meta use Qualcomm's Dragonfly C1000 CPU?
Qualcomm said the first-generation Dragonfly C1000 will be in production starting in the second half of 2028. The companies did not disclose volume commitments or pricing.
What is Modular and why is Qualcomm buying it?
Modular builds AI software meant to run models across CPUs, GPUs, NPUs, and custom ASICs without rewriting code for each chip. Reuters calculated the all-stock deal at $3.92 billion.
What is Qualcomm High Bandwidth Compute?
High Bandwidth Compute is Qualcomm's near-memory computing design for AI inference. The company says AI250 with first-generation HBC should sample in mid-2027, with AI300 and second-generation HBC expected in 2028.
Why does the 2029 target matter?
Qualcomm says handsets should fall to about one-third of QCT revenue by fiscal 2029. The data center target is the clearest number attached to its effort to build a second growth engine beyond phones.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
