Alphabet's Google collaborates with Meta to enhance the compatibility of its custom AI chips, Tensor Processing Units (TPUs), with Meta’s open-source AI software, PyTorch.
Named TorchTPU, the initiative strategically positions Google’s TPUs as a viable alternative to Nvidia’s dominant GPUs. Nvidia’s hardware currently serves as the standard for training and running most AI models.
The effort aims to break down the software barrier favoring Nvidia. Nvidia’s CUDA software platform created a strong ecosystem around its hardware.
By increasing TPU accessibility for the large community of PyTorch developers, Google seeks to capture a larger share of the booming AI infrastructure market. This collaboration underscores increasing efforts by major tech companies to reduce reliance on Nvidia and diversify the AI chip landscape.