For years, artificial intelligence has leaned heavily on Nvidias graphics processing units, largely because of CUDA. CUDA isnt just software; its a toolkit that makes developing and running AI models far simpler on Nvidia chips. The catch is that once software is written for CUDA, it wont run elsewhere. That means companies looking to scale AI systems often have no real alternative but to keep buying Nvidia hardware, which has given the company near monopoly like control. Modular wants to bre
