Next Article
Google teams up with Meta to make AI hardware more developer-friendly
Technology
Google is working on TorchTPU, a new tool that aims to enable its custom AI chips (TPUs) to work smoothly with PyTorch—the go-to platform for many AI developers.
The goal? Make it easier for people to use Google's chips instead of always relying on NVIDIA's CUDA tech, and help bridge the gap between Google's Jax and PyTorch communities.
Why this matters and what's next
Google is working closely with Meta (the folks behind PyTorch) to speed things up, and they might even open-source some parts to attract more developers.
Since 2022, Google has been selling TPUs outside its own company, giving businesses more options beyond NVIDIA when building AI projects.
This move could shake up the AI hardware scene by making it less of a one-brand show.