Modular

Qualcomm's $3.92 billion acquisition of Modular — maker of the Mojo language and MAX inference engine — is not a chip deal. It's a direct attack on CUDA, the software platform that has locked 4 million developers and their enterprises into Nvidia's ecosystem for nearly two decades. Combined with a reported $8-10 billion Tenstorrent acquisition, Qualcomm is assembling a $14 billion full-stack alternative for the $255 billion AI inference market. Here's how to assess your own Nvidia lock-in and plan a multi-vendor inference strategy.