Qualcomm has acquired a software-centric AI infrastructure startup to bolster the foundation layer for its data center and edge hardware in a clear challenge to Nvidia’s CUDA dominance.

Founded by Chris Lattner, the legendary software engineer behind open source compiler LLVM and the Swift programming language for Apple, Modular gives Qualcomm the means to offer a more optimized AI compute layer.

Its software stack provides optimizations spanning GPU kernels to API endpoints, providing cluster operators with highly intricate compute orchestration tools. The startup bills its offering as ‘write once, run anywhere,’ with users able to port the same model or codebase across hardware from Nvidia and AMD to Arm and Apple.

No financial terms of its deal were disclosed; however, Qualcomm said Modular’s software expertise will make it “well positioned to help customers move AI into production from device to cloud, with systems that are faster, more efficient, and easier to scale.”

“We believe the future belongs to developer-friendly, horizontal platforms that can run across diverse compute environments and give customers real choice in how and where they deploy AI,” Qualcomm president and CEO, Cristiano Amon, said. “With Modular, we’re accelerating that shift, combining our scale and energy-efficient data center technologies with an open ecosystem approach to help drive the next chapter of AI.”

Modular was founded by Lattner and Google alumnus Tim Davis in 2022. The pair sought to unify AI development and deployment, providing the software infrastructure to supercharge heterogeneous compute. Their platform has its own dedicated programming language, Mojo, which marries the ease of use and familiar syntax of Python with the high-performance capabilities and low-level control of languages like C or Rust.

The Thomas Tull-led US Innovative Technology (USIT) Fund led its Series C round last year that valued the company at $1.6 billion. Other backers include Google Ventures, General Catalyst, and Greylock Ventures.

Modular joins Qualcomm as it shifts from being a mobile-first chipmaker to data center AI infrastructure provider, having withdrawn from the market back in 2018.

A hint at a return emerged last summer, when Amon told investors data centers “represents a new growth opportunity for Qualcomm and is a logical extension of our diversification strategy.” He went one step further at this year’s Mobile World Congress, casting 6G from its core telecom efforts as nothing less than a global “AI data center network,” noting that future telecom infrastructure would power AI agents across devices.

More strategically, beyond mere CEO comments, Qualcomm acquired Alphawave IP Group with plans to tap its data center design capabilities set to support its market pivot.

“Modular was founded on the belief that AI needs a more open and efficient software foundation that can span diverse hardware and deployment environments,” Lattner said. “Joining Qualcomm gives us the scale and platform reach to accelerate that mission. Together, we can make AI development more accessible and performant for developers, strengthen portability across hardware, and help grow an open ecosystem that broadens participation and speeds innovation.”

Qualcomm’s data center refocus has already seen it secure hyperscale interest, with Meta revealed to be among the early customers for its forthcoming Dragonfly C1000 CPU. The chip line is set to start production in 2028, with Qualcomm positioning it as an agentic AI-optimized processor – joining the growing ranks of rival hardware, such as Arm’s AGI and Nvidia’s acqui-hired Groq line.