Qualcomm has secured a major endorsement in its quest to break into the data centre server market: Meta Platforms, the parent company of Facebook, has signed on as the first named customer for Qualcomm's upcoming Dragonfly C1000 processor. The announcement was made during Qualcomm's investor day in New York on Wednesday, alongside the unveiling of a new AI300 accelerator chip and a confirmed $3.9 billion all-stock acquisition of AI software startup Modular.
The Dragonfly C1000 is a general-purpose server CPU designed to sit inside data centres alongside Qualcomm's AI accelerator chips. According to Qualcomm, Meta has committed to using the C1000 and its successors across its facilities. However, the chip is not expected to become available until 2028, meaning the partnership is a forward-looking commitment rather than an immediate deployment. This timeline gives Qualcomm years to refine the architecture and prove its power-efficiency claims — a central selling point for the company that has long dominated the mobile chip market.
Dragonfly brand expands Qualcomm's data centre ambitions
The Dragonfly brand was first introduced by Qualcomm at Computex in early June, alongside an ASIC supply deal with ByteDance. The brand covers three product categories: data centre CPUs, AI inference accelerators, and custom silicon built in collaboration with hyperscale cloud providers. Wednesday's investor day filled in many of the product details that the Computex teaser had left out, offering a more complete picture of Qualcomm's strategy.
Qualcomm's data centre push is not its first. The company previously attempted to enter the server market in 2017 with the Centriq processor, a 64-bit ARM-based chip aimed at cloud providers. That effort was ultimately shut down in 2018 after failing to gain traction, partly because of limited software ecosystem support and the dominance of Intel's Xeon processors. This time, however, the company is approaching the market with a different strategy: a three-pronged portfolio of CPUs, AI accelerators, and custom chips, combined with a software acquisition that addresses the critical developer lock-in problem.
The custom silicon category is particularly noteworthy. By offering to build bespoke chips for hyperscalers like ByteDance and potentially others, Qualcomm is positioning itself as a flexible partner that can meet specific workload requirements. This mirrors a broader industry trend where cloud giants such as Amazon, Google, and Microsoft design their own processors for internal use, reducing reliance on traditional vendors like Intel and AMD. Qualcomm hopes that its expertise in mobile SoCs, particularly in power efficiency and integrated design, will give it an edge in the data centre.
AI accelerator lineup expands with AI300
On the accelerator side, Qualcomm added the AI300 chip to a lineup that already included the AI200 and AI250. The AI200, built on Qualcomm's Hexagon neural processing unit technology, features direct liquid cooling and supports up to 768GB of LPDDR memory. It is on track for initial customer shipments later this year. The AI250 is expected to follow in 2027. The newly announced AI300 sits above these chips in the portfolio, though specific performance benchmarks were not disclosed.
These accelerators are designed for inference, the process of running trained AI models at scale rather than training them from scratch. Qualcomm argues that its decades of mobile chip design give it a distinct advantage in power efficiency. As data centres strain electricity grids worldwide, any power savings can translate directly into lower operating costs and a smaller carbon footprint. However, whether that mobile expertise translates to data centre performance remains unproven at scale. The inference market is already crowded, with Nvidia's GPUs leading, AMD's Instinct accelerators gaining ground, and custom ASICs from Google, Amazon, and others offering tailored solutions.
Qualcomm's inference strategy is also notable because it targets a growing segment of the AI market. As large language models and generative AI applications move from training to deployment, the demand for efficient inference hardware is skyrocketing. Meta, for example, runs inference for its social media platforms, internal AI tools, and the metaverse. By securing Meta as a customer, Qualcomm gains a high-profile advocate that can validate its technology in real-world deployments.
Modular acquisition: software as the key to unlocking hardware value
The Modular acquisition, which TNW reported was nearing completion on Monday, is a critical piece of Qualcomm's strategy. Modular is the company behind the Mojo programming language and the MAX inference engine. MAX is a cross-platform inference engine that allows AI models to run efficiently on chips from Nvidia, AMD, Intel, and Qualcomm without developers having to rewrite code for each processor. This is a direct challenge to Nvidia's CUDA platform, the proprietary software layer that has locked AI developers into Nvidia hardware for nearly two decades.
Breaking that lock-in is the central challenge for every company trying to compete with Nvidia in AI infrastructure. Nvidia's GPUs are powerful, but the real moat is CUDA: a rich ecosystem of libraries, tools, and community code that makes it easier for developers to build and deploy models on Nvidia hardware. Without a comparable software ecosystem, even the best hardware struggles to gain adoption. Qualcomm's acquisition of Modular addresses this head-on.
CEO Cristiano Amon framed the deal as part of an industry movement toward open, multi-vendor architectures. In his investor day presentation, Amon positioned Qualcomm as the anti-Nvidia, offering flexibility where Nvidia's CUDA demands loyalty. The total cost of the acquisition is roughly $3.9 billion, paid entirely in stock. Qualcomm will issue about 19 million shares to Modular's owners, illustrating the high value the company places on software talent and intellectual property. The deal is expected to close in the second half of 2025.
The strategic logic is straightforward. Qualcomm can design competitive chips, but without a software ecosystem that makes developers want to use them, the hardware alone is insufficient. Modular's cross-platform tooling could give Qualcomm the kind of developer on-ramp it currently lacks. Moreover, Modular's Mojo language is designed to combine the performance of low-level programming with the productivity of Python, making it attractive for AI researchers and engineers who want to write efficient code without sacrificing ease of use.
Meta partnership: diversification signal for hyperscaler AI infrastructure
The Meta partnership is notable for what it implies about diversification in the AI hardware supply chain. Meta currently builds its AI infrastructure primarily around Nvidia GPUs, and has also invested in its own custom MTIA chips for inference. Adding Qualcomm to that mix suggests Meta wants more supplier options as it scales inference, not that it is replacing Nvidia. In fact, Nvidia announced a multiyear strategic partnership with Meta earlier this year, further deepening their existing relationship.
For Qualcomm, having a marquee customer like Meta provides several advantages. First, it validates the company's technology direction and gives investors confidence that there is real demand. Second, it gives Qualcomm a design partner that can provide feedback on architecture and performance needs. Third, it creates a reference implementation that Qualcomm can use to win other hyperscaler customers, such as Google, Amazon, or Microsoft.
The deal also highlights the shift in AI workload mix. Training large models still requires massive clusters of Nvidia H100 or B200 GPUs, but as models are deployed, inference becomes the dominant cost. Meta, with billions of users, runs inference on virtually every interaction on its platforms. If Qualcomm's chips can deliver comparable performance with lower power consumption, Meta could significantly reduce its operational costs. However, significant integration work remains: the C1000 will not arrive until 2028, and software compatibility with Meta's existing stack needs to be ensured.
Qualcomm's financial outlook and competitive landscape
Qualcomm shares have climbed about 30 percent in 2025 on expectations that AI would open a second growth engine beyond smartphones, which still account for the majority of the company's revenue. The investor day was designed to turn that expectation into a concrete roadmap. With the Modular acquisition providing the software layer, Meta providing the first marquee customer, and the AI200 approaching shipments, the pieces are assembling on paper.
Whether they assemble in practice depends on execution over the next two to three years. The C1000 does not ship until 2028, meaning Qualcomm will need to convince hyperscalers to commit to a product that won't be available for several years. The Modular deal has not yet closed, and integrating the company's culture and technology into Qualcomm will take time. The AI accelerator lineup has no published benchmarks against Nvidia's current or upcoming hardware, so performance claims remain unverified.
Moreover, the competitive landscape is fierce. Nvidia's next-generation Blackwell architecture is already shipping, and the company continues to invest heavily in both hardware and software. AMD is gaining traction with its Instinct MI300 series, and Intel is working on its Gaudi accelerators. Meanwhile, every major cloud provider is designing custom silicon: Amazon's Trainium and Inferentia, Google's TPU, and Microsoft's Maia. Qualcomm may be entering a race that is already won by incumbents with deep pockets and established relationships.
However, Qualcomm does have unique strengths. Its deep experience in power-efficient design, honed over decades of supplying chips for smartphones, could become a differentiator as data centres face growing energy constraints. The company also has a massive balance sheet, strong relationships with contract manufacturers like TSMC, and a broad portfolio of wireless and connectivity technologies that could be integrated into its data centre offerings. Additionally, by offering a range of products from CPUs to accelerators to custom ASICs, Qualcomm can serve as a one-stop shop for hyperscalers looking to diversify their supply chains.
The announcement of the Dragonfly C1000 and the Meta partnership marks a significant milestone for Qualcomm, but the road ahead is long. The company must deliver on its technical promises, win additional customers, and navigate a rapidly evolving market. For now, the pieces are in place, but the real test will come when the chips are in production and the benchmarks are public.