Qualcomm secures Meta as its inaugural named customer for its Dragonfly data center chips.

      Qualcomm has officially selected Meta as its first customer for the Dragonfly C1000 data centre chip, which is set for release in 2028. Additionally, Qualcomm confirmed a $3.9 billion acquisition of AI software startup Modular.

      The announcement was made during Qualcomm's investor day in New York, showcasing a commitment to competing in the AI infrastructure sector. Along with the Dragonfly C1000, the company unveiled a new AI300 accelerator chip and provided details on the Modular acquisition.

      The Dragonfly C1000 is a versatile server processor intended for data centres, and Meta plans to utilize this chip and its future versions throughout its operations. However, since the chip won’t be available until 2028, this partnership represents a long-term agreement rather than something that will be deployed immediately.

      Qualcomm's Dragonfly brand, introduced at Computex in June, includes data centre CPUs, AI inference accelerators, and custom silicon for hyperscalers. The latest event clarified details regarding these products.

      On the accelerator front, Qualcomm introduced the AI300 chip, expanding its AI lineup that already includes the AI200 and AI250. The AI200, leveraging Qualcomm’s Hexagon neural processing unit technology and offering up to 768GB of LPDDR memory, is scheduled for initial customer shipments later this year, while the AI250 is expected to launch in 2027.

      These accelerators focus on inference, which involves running AI models that have already been trained, rather than creating them anew. Qualcomm believes its long experience with mobile chip design will enhance power efficiency, particularly as global data centres face increasing electricity demands. However, it remains to be seen if this mobile expertise can translate effectively to data centre performance at scale.

      The Modular acquisition, which was reported to be nearing completion earlier this week, is now confirmed to involve about four billion dollars in stock. Qualcomm will issue about 19 million shares to Modular’s owners, and the transaction is anticipated to close in the latter half of this year.

      Modular develops the Mojo programming language and the MAX inference engine, allowing AI models to operate across multiple chip platforms from Nvidia, AMD, Intel, and Qualcomm without requiring developers to rewrite code for each specific processor. This directly challenges Nvidia’s CUDA platform, which has kept AI developers tied to Nvidia hardware for the past two decades. Overcoming this software dependency is a major hurdle for any company aiming to compete with Nvidia in the AI infrastructure space.

      Qualcomm’s strategy is clear: while it can produce competitive hardware, without a robust software ecosystem to encourage developer adoption, hardware alone won’t suffice. Modular’s cross-platform tools could provide Qualcomm with a much-needed entry point for developers.

      CEO Cristiano Amon positioned the acquisition as part of a broader industry trend towards open, multi-vendor architectures, casting Qualcomm as an alternative to Nvidia by offering flexibility in contrast to Nvidia’s reliance on loyalty via CUDA.

      Although Qualcomm’s ambitions are high, its experience in the data centre market is limited. Most of its revenue comes from smartphone processors and modems, and its earlier attempt to penetrate the server market with the Centriq processor in 2017 was unsuccessful. The current initiative, however, benefits from greater institutional backing, a confirmed hyperscaler client in Meta, and a clearer opportunity in AI inference, yet there remains a significant gap between announcements and actual revenue.

      The collaboration with Meta is noteworthy for what it suggests about diversification. Currently, Meta's AI infrastructure mainly relies on Nvidia GPUs, along with its custom MTIA chips. By integrating Qualcomm into its strategy, Meta appears to seek a broader range of supplier options for scaling inference, rather than eliminating Nvidia, which recently announced a multi-year strategic partnership with the company.

      Qualcomm’s stock has risen around 30 percent this year, driven by expectations that AI will provide a new growth avenue beyond smartphones. The investor day aimed to transform these expectations into a concrete plan. With Modular adding a software dimension, Meta serving as the first key customer, and the AI200 nearing shipment, many elements seem to be aligning on paper.

      However, the realization of this potential will depend on effective execution over the next two years. The C1000 won’t be released until 2028, the Modular acquisition is still pending completion, and the AI accelerator range lacks published benchmarks against Nvidia’s existing or forthcoming hardware. Qualcomm is making strategic moves to enter this market, but it is stepping into a highly competitive arena where Nvidia currently holds a substantial lead and all major cloud providers are also developing custom silicon.