Meta uses Google’s AI chips for its AI workloads

Meta is now using Google’s tensor processing units for its AI workloads. This new version of the SGLang inference engine allows Meta to run its AI workloads on Google’s TPUs without rewriting much of its code. This is the first time Google has sold its TPUs to external customers, and it could challenge Nvidia’s market dominance. Google’s TPUs are now running in Meta’s data centers, and this is the latest development in the use of AI chips for AI workloads.

Meta’s PyTorch-TPU lead, Zongwei Zhou, said that the new version of SGLang gives Meta a remarkably mature, PyTorch-native serving stack. This is the result of a close three-way collaboration between Meta, Google, and the SGLang team. Meta’s PyTorch engineering director, Jana van Greunen, added that Meta cannot be constrained by a single hardware architecture, and that TorchTPU enables it to deploy its PyTorch workloads to Google TPUs. This is not the first time Meta has used Google’s TPUs.

This is the first time Google has sold its TPUs to external customers. This could challenge Nvidia’s market dominance. Google’s TPUs are now running in Meta’s data centers, and this is the latest development in the use of AI chips for AI workloads.

This is the result of a close three-way collaboration between Meta, Google, and the SGLang team. Meta’s PyTorch engineering director, Jana van Greunen, added that Meta cannot be constrained by a single hardware architecture, and that TorchTPU enables it to deploy its PyTorch workloads to Google TPUs. This is not the first time Meta has used Google’s TPUs.

This is the first time Google has sold its TPUs to external customers. This could challenge Nvidia’s market dominance. Google’s TPUs are now running in Meta’s data centers, and this is the latest development in the use of AI chips for AI workloads.