Google, Amazon challenge Nvidia’s GPU lock-in

Google and Amazon are both trying to loosen Nvidia’s grip on AI infrastructure by making it easier for PyTorch developers to move applications off Nvidia GPUs, according to TechCrunch. The two cloud giants have released developer tools aimed at reducing switching costs, a direct challenge to the CUDA software ecosystem that has helped Nvidia dominate AI computing. Nvidia’s advantage is not only hardware performance; it is also the maturity of its software stack, libraries, and developer familiarity. That ecosystem creates friction for customers considering alternatives. By offering tools that simplify porting PyTorch workloads, Google and Amazon hope to steer more AI training and inference onto their own chips and cloud platforms. For customers, the appeal is clear: less dependence on scarce or expensive Nvidia hardware, and potentially more bargaining power. But adoption will depend on real-world performance, compatibility, support, and the effort required to rewrite or optimize workloads. The move does not mean Nvidia’s position will collapse overnight. It does signal that major cloud providers want to commoditize the GPU layer and capture more of the value created by AI workloads. If the tools mature, they could give developers a practical path to test alternatives and push the AI chip market toward more competition.