Alarm for Nvidia; Google reduces CUDA dependency with TorchTPU project

Google With a new project called TorchTPUalong with Meta It tries to improve the compatibility of its artificial intelligence processors, so-called TPUwith the framework PyTorch to increase The main goal of this project is to reduce the dependence of artificial intelligence developers on the software ecosystem Nvidia CUDA It has been stated.

According to Reuters, Google is working on a plan that will allow the world’s most widely used open source framework for developing artificial intelligence models, PyTorch, to run better on TPUs. PyTorch, of which Meta is one of the biggest supporters, is one of the key tools for AI developers. Developers use such frameworks to automate many different tasks.

According to knowledgeable sources, Google is even considering open source part of this software to speed up its acceptance process for customers.

Without going into project details, a Google Cloud spokesperson confirmed to Reuters that the move is aimed at increasing customer choice and that demand for both TPU and GPU infrastructure is growing rapidly.

More serious competition between Google and Nvidia in the artificial intelligence computing market

This plan directly Nvidia position It targets artificial intelligence in the computing market. In recent years, Nvidia has been able to dominate the market by relying on the combination of powerful hardware and its CUDA software ecosystem. One of the most important factors of the superiority of Nvidia’s CUDA software ecosystem against competitors is PyTorch support.

Reuters also says Meta, which is looking to reduce inference costs and diversify its AI infrastructure, will benefit from the partnership by reducing its reliance on Nvidia. At the same time, Google has also started direct sales of TPU to customers’ data centers and by appointing “Amin Vahdat” as the head of artificial intelligence infrastructure, it has shown its determination to compete more seriously in this market.

The sale of Google TPUs in the past years has become one of the most important reasons for the growth of the company’s cloud sector (Google Cloud) and the technology giant is trying to show investors that its investments in the field of artificial intelligence are accompanied by economic returns by introducing these programs.