Qualcomm names Meta as first customer for Dragonfly chips
▼ Summary
– Qualcomm signed Meta as the first named customer for its Dragonfly C1000 data centre processor, a general-purpose server chip that will not be available until 2028.
– Qualcomm confirmed the acquisition of AI software startup Modular for roughly $3.9 billion in stock, gaining the Mojo programming language and MAX inference engine.
– The company announced a new AI300 accelerator chip, joining the AI200 (shipping later this year) and AI250 (expected in 2027), all designed for AI inference.
– The Modular acquisition gives Qualcomm cross-platform software that challenges Nvidia’s CUDA lock-in, allowing AI models to run on multiple chip brands without rewriting code.
– Qualcomm’s data centre track record is thin, as its previous Centriq server chip failed in 2017, and the C1000’s success depends on execution against Nvidia’s commanding lead.
Qualcomm has officially named Meta as the first customer for its upcoming Dragonfly C1000 data centre processor, marking a pivotal step in the mobile chipmaker’s push into the fiercely competitive AI infrastructure market. The announcement came during Qualcomm’s investor day in New York on Wednesday, where the company also unveiled a new AI300 accelerator chip and confirmed the acquisition of AI software startup Modular for approximately $3.9 billion in stock.
The Dragonfly C1000 is a general-purpose server processor built to work alongside Qualcomm’s AI accelerators inside data centres. Meta has committed to deploying the C1000 and its future iterations across its facilities. However, the chip will not be available until 2028, making this a long-term strategic pledge rather than an immediate rollout.
Qualcomm first introduced the Dragonfly brand at Computex in early June, alongside an ASIC supply agreement with ByteDance. The brand now covers three distinct product lines: data centre CPUs, AI inference accelerators, and custom silicon co-developed with hyperscalers. Wednesday’s event fleshed out the product specifics that were left vague at Computex.
On the accelerator front, Qualcomm expanded its lineup with the AI300, joining the previously announced AI200 and AI250. The AI200, which leverages Qualcomm’s Hexagon neural processing unit technology and features direct liquid cooling with up to 768GB of LPDDR memory, is scheduled for initial customer shipments later this year. The AI250 is expected to follow in 2027.
These chips are designed specifically for inference,the process of running trained AI models at scale, as opposed to training them from scratch. Qualcomm argues that its extensive experience in mobile chip design gives it a unique advantage in power efficiency, a critical factor as data centres place increasing strain on global electricity grids. Whether that mobile expertise translates effectively to data centre workloads remains an open question.
The Modular acquisition, which TNW reported was nearing completion earlier this week, is now confirmed as an all-stock transaction valued at roughly $4 billion. Qualcomm will issue approximately 19 million shares to Modular’s owners. The deal is expected to close in the second half of this year.
Modular is best known for developing the Mojo programming language and the MAX inference engine, software that enables AI models to run across chips from Nvidia, AMD, Intel, and Qualcomm without developers needing to rewrite code for each platform. This capability directly challenges Nvidia’s CUDA platform, the software ecosystem that has locked AI developers into Nvidia hardware for nearly two decades. Breaking that lock-in is the central hurdle for any company aiming to compete with Nvidia in AI infrastructure.
The strategic logic behind the acquisition is clear. Qualcomm can design competitive chips, but without a robust software ecosystem to attract developers, the hardware alone won’t suffice. Modular’s cross-platform tooling could provide the developer on-ramp that Qualcomm currently lacks.
CEO Cristiano Amon framed the deal as part of a broader industry shift toward open, multi-vendor architectures. This positioning casts Qualcomm as the anti-Nvidia, offering flexibility where Nvidia’s CUDA demands loyalty.
Qualcomm’s ambitions are substantial, but its track record in the data centre space is limited. The company generates the vast majority of its revenue from smartphone processors and modems. Its previous attempt to enter the server market with the Centriq processor in 2017 ended in failure. The current effort enjoys more institutional support, a named hyperscaler customer in Meta, and a clearer market opportunity in AI inference. Still, the gap between investor day announcements and actual shipped revenue remains wide.
The Meta partnership is particularly significant for what it signals about diversification. Meta currently builds its AI infrastructure primarily around Nvidia GPUs and has also invested in its own custom MTIA chips. Adding Qualcomm to the mix suggests Meta wants more supplier options as it scales inference operations, not that it is replacing Nvidia, which announced a multiyear strategic partnership with Meta earlier this year.
Qualcomm shares have risen roughly 30 percent this year on expectations that AI will unlock a second growth engine beyond smartphones. The investor day was designed to turn that expectation into a concrete roadmap. With the Modular acquisition providing the software layer, Meta serving as the first marquee customer, and the AI200 approaching shipments, the pieces are coming together on paper.
Whether they come together in practice depends on execution over the next two years. The C1000 won’t ship until 2028, the Modular deal hasn’t closed, and the AI accelerator lineup lacks published benchmarks against Nvidia’s current or upcoming hardware. Qualcomm is making the right moves to enter the market, but it is entering a race where Nvidia holds a commanding lead and every major cloud provider is also designing custom silicon.
(Source: The Next Web)
