From TechCrunch · Julie Bort · · 1m 30s
Google and Amazon challenge Nvidia's AI chip dominance
Google and Amazon release PyTorch versions for their own chips, trying to break Nvidia's grip on AI development.
In brief
Google and Amazon release PyTorch versions for their own chips, trying to break Nvidia's grip on AI development. Google and Amazon released PyTorch versions for their own AI chips, letting developers switch from Nvidia hardware without rewriting code. Originally reported by TechCrunch.
Quick answers
- The real test: will developers switch?
- Amazon and Google are clearly working to end Nvidia's advantage. But AI app developers will not automatically ditch Nvidia's hardware just because they can now easily move to alternative chips and clouds.
- What is the key takeaway from Google and Amazon challenge Nvidia's AI chip dominance?
- Google and Amazon released PyTorch versions for their own AI chips, letting developers switch from Nvidia hardware without rewriting code.
Google launches PyTorch for its chips
On Tuesday, Google announced TorchTPU, an open source version of PyTorch that runs on Google's Tensor Processing Units. It lets developers move existing PyTorch apps to TPUs with just a configuration change.
Why this matters: PyTorch dominates AI
PyTorch is the popular open source toolkit used by eighty five percent of AI apps. It was originally developed by Meta in two thousand sixteen and is now managed by the Linux Foundation.
Amazon makes a similar move
Amazon released TorchNeuron on Monday, its own PyTorch version for its Trainium chip and AWS. Amazon is also submitting code to the main PyTorch project to decouple features from Nvidia's chips.
Google goes further with developer tools
Google worked with Meta, Hugging Face, Inferact, RadixArk and others to ensure TorchTPU works with popular AI development tools. Google also launched the TPU agent suite to help developers move apps and fix errors.
The real test: will developers switch?
Amazon and Google are clearly working to end Nvidia's advantage. But AI app developers will not automatically ditch Nvidia's hardware just because they can now easily move to alternative chips and clouds.
What to remember
Google and Amazon released PyTorch versions for their own AI chips, letting developers switch from Nvidia hardware without rewriting code.
