宝玉 (@dotey) · 10/21/2026, 14:09:15
Meta Confirms Production Use of Google TPUs for AI Inference
Meta engineers confirmed that SGLang-TorchTPU is running in production at Meta, marking the formal adoption of Google TPUs for AI inference. This follows reports of a multi-billion dollar, multi-year TPU leasing deal between Meta and Google, posing a potential challenge to NVIDIA's CUDA ecosystem.
SOURCE COVERAGEFull Original Report
Meta Engineers Confirm: Meta Is Now Running AI Inference on Google TPUs in Production
On October 20, Google published a blog post introducing the TPU version of the SGLang inference engine. The article quoted Zongwei Zhou, Head of PyTorch-TPU at Meta, stating: "SGLang-TorchTPU is already running in Meta’s production environment today." This remains the only first-hand confirmation to date.
TPUs are Google’s custom-designed AI chips, competing directly with NVIDIA GPUs. Inference refers to the process where an AI model performs its actual work—every time you ask a question and receive an answer, that computation constitutes one inference step.
Zongwei Zhou did not disclose the scale of deployment, which specific products utilize this setup, or whether Meta leases or owns the TPUs. However, The Information reported in February that Meta and Google had signed a multi-year, multi-billion-dollar contract for Meta to lease TPUs, though neither company has publicly confirmed the details.
Google Developer Blog: https://t.co/1pKn6zeugg PyTorch Blog: https://t.co/9MglhcELsd
↳ Quoting @SemiAnalysis_ (SemiAnalysis): BREAKING NEWS🚨🚨: TODAY META CONFIRMS THEY ARE RUNNING TPU INFERENCE IN PRODUCTION. What does this mean for the CUDA moat?
For those who read our public newsletter, this is not a surprise, as on Nov 28, 2025, we publicly wrote that Meta is a TPU customer & has lots of engineers working on TPUs. Our accelerator supply chain model has details about Meta's TPU procurement!
