Qualcomm has agreed to acquire AI software startup Modular for roughly $3.9 billion in an all-stock deal, the company confirmed at its Investor Day on June 24, 2026. The purchase hands Qualcomm the Mojo programming language and MAX inference engine, a software layer designed to let AI models run across any chip, in a direct strike at NVIDIA's CUDA moat.
Buying software, not silicon
Rather than adding raw compute, Qualcomm is acquiring what it calls "a silicon-agnostic compute layer." Modular's MAX engine and Mojo language allow developers to run models across NVIDIA, AMD, Intel, ARM and Qualcomm hardware without rewriting code, the kind of portability that could lower the switching costs keeping workloads pinned to NVIDIA's ecosystem. Qualcomm will issue about 19.2 million shares for the company, with the deal expected to close in the second half of 2026.
The Lattner factor
The acquisition brings roughly 150 employees, including Modular co-founders Chris Lattner, the creator of LLVM and Swift, and Tim Davis. Analysts noted that the real value lies less in technology than in talent: a credible "write once, run across CPU, GPU, NPU and ASIC" layer is exactly what makes non-NVIDIA silicon a safer enterprise bet. Even so, skeptics cautioned that CUDA's decade-deep moat will take years, not months, to erode.
Part of a $14 billion AI stack
The Modular deal lands alongside Qualcomm's reported pursuit of Tenstorrent, the RISC-V AI chip startup led by Jim Keller. If both transactions close, Qualcomm will have spent up to $14 billion assembling an integrated AI stack spanning custom silicon, software and its Snapdragon and XPU accelerator roadmap. The move continues a wave of consolidation in AI infrastructure that has included NVIDIA's purchase of Kumo AI and AMD's acquisition of Mext.
Reporting based on coverage from Bloomberg, Network World and Qualcomm.