Qualcomm has outlined how its approximately $4 billion acquisition of AI software company Modular will support a unified development layer across different AI chips. The strategy, presented at Snapdragon Summit 2026 in Maui, Hawaii, is designed to reduce separate development and optimization work for central processing units, graphics processing units and neural processing units. Modular’s Mojo programming language, MAX model-runtime system and Modular Cloud infrastructure platform currently support six architectures from six major hardware vendors in one environment. Qualcomm plans to extend the software from its Dragonwing data-center AI accelerator to Snapdragon chips used in PCs and smartphones. Microsoft is working with the companies to support Mojo on Windows without separate conversion. The effort positions Qualcomm against NVIDIA’s CUDA ecosystem and shifts competition toward software and developer adoption, although Modular’s Snapdragon integration remains at an early stage.