Meta now runs AI workloads on Google TPUs: Google’s TPUs now power Meta’s AI workloads, marking a new chapter in AI development.

Meta is now running AI workloads on Google TPUs through a new PyTorch-native version of the SGLang inference engine. This development comes after a close three-way collaboration between Meta, Google, and the SGLang team. Meta’s PyTorch-TPU lead Zongwei Zhou said the new versions give Meta a remarkably mature, PyTorch-native serving stack. This means Meta can now deploy its PyTorch workloads on Google TPUs without being constrained by a single hardware architecture.

Google and Meta co-developed TorchTPU, which is the new native option for PyTorch. This move is part of a broader effort to loosen Nvidia’s software lock-in. Amazon also open-sourced its own PyTorch backend for its Trainium2 and Trainium3 chips, adding to the competition in the AI chip market.

Although it is unclear where Meta’s production TPUs sit, talks last year included Google Cloud TPU rentals in 2026 with purchases in 2027. Meta later signed a multi-year rental agreement for TPUs. This announcement came just a day after Amazon’s move, showing that both companies are aiming to break Nvidia’s dominance in the AI chip market.

A tuned vLLM setup serving Moonshot AI’s open-weight, 2.8-trillion-parameter Kimi K3 model reportedly tops 700 tokens per second for a single user on the seventh-generation Ironwood TPU. This speed figure comes from Inferact, the startup founded by vLLM’s creators that helped Google build vLLM-TorchTPU. Its test ran 16 Ironwood chips with speculative decoding, a method of speeding up large language model inference with a small draft model, against vLLM’s published setup on 16 Nvidia GB200s, which reached 452 tokens per second.

SGLang-TorchTPU will keep adding models, starting with DeepSeek V4.1 Flash and Reflection Beam. It's also working toward Day 0 TPU support for new open models, meaning support on release day, by the first quarter of 2027, according to RadixArk. Google, Meta and RadixArk plan to share more at SGLang Summit 2026, Nov. 12-13 in San Francisco.