Qualcomm Bets $14 Billion on Breaking Nvidia's Data Center Lock
At its June 24 Investor Day, Qualcomm confirmed a $3.9 billion acquisition of AI software startup Modular and disclosed ongoing talks to buy chip designer Tenstorrent for up to $10 billion, a two-pronged push to build an Nvidia alternative from the silicon up.
Qualcomm used its June 24 Investor Day in New York to confirm what it had been assembling in pieces for two years: a full-stack data center architecture meant to give cloud providers a credible alternative to Nvidia hardware.
The confirmed piece is Modular. Qualcomm announced a $3.92 billion all-stock deal to acquire the AI software startup, with the transaction expected to close in the second half of 2026, subject to regulatory approval, according to Reuters calculations cited in reporting by CNBC. Modular builds a hardware-agnostic inference layer, software that lets AI models run across CPUs, GPUs, NPUs, and custom accelerators without requiring developers to rewrite code for each processor. That's the specific problem Qualcomm has struggled with for years. ServeTheHome, whose reporter attended the Investor Day in person, noted that the existing Qualcomm AI100 accelerator has a software stack that is, charitably, hard to use. Modular is the fix for that.
The unconfirmed piece is Tenstorrent. As reported by The Information on June 15 and confirmed as a reported development by Reuters, Qualcomm is in talks to acquire the AI chip startup at a valuation of $8 billion to $10 billion. Talks are ongoing and no deal has been announced. Qualcomm told The Register it doesn't comment on rumors. Tenstorrent declined to comment to The Information.
What makes the Tenstorrent talks architecturally interesting is what that company actually builds. Tenstorrent designs AI accelerators on the open RISC-V instruction set standard, and its Blackhole chip reached general availability in April 2026. The company's Series D in late 2024, which valued it at $2.6 billion and drew investment from Bezos Expeditions and Samsung Securities, was followed by reported talks to raise at a $3.2 billion valuation in 2025. A deal at $8 to $10 billion would represent a sharp step up, and whether the final price includes performance-based milestone payments remains unclear, according to Reuters.
If both deals close, Qualcomm will have committed over $14 billion to a single strategic goal, per TechTimes reporting: making it possible to run AI inference workloads on hardware that doesn't come from Nvidia. The logic of the stack is readable from the parts. Qualcomm's earlier $2.4 billion acquisition of Alphawave Semi covers high-speed connectivity between chips in an AI cluster. The December 2025 acquisition of Ventana Micro Systems adds a RISC-V CPU for server chiplets. Tenstorrent, if it closes, adds the accelerator. Modular ties the software layer across all of it.
Also announced at the Investor Day: Meta has signed a multi-generational agreement to deploy Qualcomm's Dragonfly C1000 server CPU, a 250-plus-core Arm-based chip, in its next-generation server fleet, according to TechTimes. Microsoft was named as a second anchor partner. The C1000 won't start production until the second half of 2028, which is where the execution question lives.
Qualcomm CEO Cristiano Amon set a target of $15 billion in annual data center AI chip sales by fiscal 2029, up from $5 billion expected in 2027, according to Yahoo Finance. Qualcomm shares jumped more than 13% in after-hours trading following the Investor Day disclosures, then gave most of it back by the close on June 25.
The skeptic case isn't complicated. The Dragonfly C1000 ships in 2028. The AI accelerators are still being designed. Analysts at Bank of America, who maintained an Underperform rating ahead of the Investor Day, flagged execution risk and the difficulty of challenging Nvidia's CUDA software ecosystem, according to TradingKey reporting. The Modular acquisition is precisely the answer to the CUDA argument, a compiler stack that doesn't need CUDA, but it's a software company acquiring the customers and trust that Nvidia has spent fifteen years building. That doesn't close in the second half of 2026.
Sources cited:
- CNBC (https://www.cnbc.com/2026/06/24/qualcomm-ai-chip-modular-software.html)
- The Register (https://www.theregister.com/systems/2026/06/16/qualcomm-said-to-be-circling-ai-chip-biz-tenstorrent-in-10b-risc-v-power-play/5256084)
- TechTimes (https://www.techtimes.com/articles/319166/20260626/qualcomm-closes-39-billion-modular-deal-meta-validates-full-stack-cuda-challenge.htm)
- Yahoo Finance / Reuters (https://finance.yahoo.com/technology/ai/articles/qualcomm-talks-acquire-ai-chip-230401789.html)
- ServeTheHome (https://www.servethehome.com/qualcomm-investor-day-2026-data-center-announcements-cpus-ai-accelerators-and-more/)
- Data Center Dynamics (https://www.datacenterdynamics.com/en/news/qualcomm-considers-acquiring-ai-chip-firm-tenstorrent/)
- TradingKey (https://www.tradingkey.com/news/market-movers/261988102-market-movers-qcom-20260624)
- TechTimes (Investor Day / $14B framing) (https://www.techtimes.com/articles/319017/20260624/qualcomm-bets-14-billion-cracking-nvidias-ai-monopoly-risc-v-open-compiler.htm)
This release was originally distributed via ETL Newswire. Visit CNBC for the full story, related releases, and contact information.
Visit CNBC →