TPU (Tensor Processing Unit)
A TPU is a custom AI accelerator chip that Google designs specifically for training and running neural networks, built as an alternative to general-purpose GPUs. Google runs most of its own AI stack on TPUs, from the silicon up through models like Gemini, rather than depending on third-party GPU supply.
Also known as: tensor processing unit, tensor processing units, tpus
GPUs were built for graphics rendering and adapted for machine learning later. TPUs are ASICs designed for the matrix multiplication that dominates neural network workloads, and for very little else. Because Google controls the chip design, the compiler stack (XLA), and the models that run on top of both, it can tune the whole pipeline together in a way that is harder when you build on someone else’s hardware, which is what most AI labs do with NVIDIA.
That vertical integration is part of why Google can price its model tiers so far apart. Pairing custom silicon with smaller model architectures pushes the cost per token of inference down far enough to serve a lot of users cheaply. TPUs also underpin the large training runs, where thousands of chips are networked across data centers to train frontier models like Gemini.
From the conversation
-
Gemini 3 & Robot Dogs: Inside Google DeepMind's AI Experiments | Paige Bailey - The Making of Gemini 2.0: DeepMind's Approach to AI Development and Deployment | Logan Kilpatrick
-
The AI Framework Era Is Over: Why Context Is the Moat | Jerry Liu - Context Poisoning is Killing Your AI Agents: How to Stop it