Google has joined Amazon in attacking Nvidia’s stronghold with AI app developers
techcrunch ·
Google has officially released TorchTPU, an open-source version of PyTorch designed to run AI applications on its Tensor Processing Units (TPUs). This move aligns Google with Amazon, which recently launched TorchNeuron for its Trainium chips, as both tech giants attempt to reduce Nvidia's dominance in the AI hardware market. By enabling developers to port existing PyTorch apps to their respective clouds without rewriting code, these companies are lowering the barrier to entry for alternative infrastructure. Google claims to have collaborated directly with major AI tool providers like Meta and Hugging Face to ensure compatibility. Additionally, Google introduced a TPU agent suite and updated LiteRT to assist developers in migrating their workloads. While these initiatives make it easier to switch away from Nvidia GPUs, they do not guarantee that developers will immediately abandon Nvidia's ecosystem. The strategic push by Google and Amazon highlights a growing effort to diversify the underlying hardware used for large language models and other AI tasks.
techcrunch