TPU v7
TPU v7 is a Google TPU based on the Inference-first MXU architecture, first released in 2025.
Built from BenchGecko data as of April 14, 2026 · updates when the data changes
TPU v7 is a Google TPU based on the Inference-first MXU architecture, first released in 2025.
Basic
The Google TPU v7 Ironwood is an AI TPU made by Google. Peak throughput is 4,614 TFLOPS at FP16/BF16 and 9,228 TFLOPS at FP8. It carries 192 GB of HBM3e with 7.37 TB/s of memory bandwidth. Rated power (TDP) is 600 W. It is built on N3 at TSMC.
Deep
In the BenchGecko dataset it sits in the frontier tier: current-generation flagship silicon shipping at scale to hyperscalers. It follows the TPU v6e. Disclosed buyers include Google. Specs come from manufacturer datasheets; the spec card on this page shows the dataset values and their as-of date.
Expert
That is 7.69 FP16 TFLOPS per watt. Memory bandwidth per unit of compute is 1.6 GB/s per FP16 TFLOP: a higher ratio helps memory-bound inference, while peak TFLOPS matter more for compute-bound training. Fabrication: Taiwan. Real throughput depends on the model, precision, batch size and software stack, so compare tokens per second per dollar for your own workload.
Depending on why you're here
- ·TPU v7: 4,614 TFLOPS FP16/BF16 · 9,228 TFLOPS FP8
- ·Process: N3 at TSMC
- ·Tier: frontier
- ·192 GB of HBM3e per chip
- ·Cloud availability in the dataset: Google Cloud
- ·Compare it with other chips on /hardware/tpu-v7
- ·Disclosed buyers: Google
- ·Google · fabbed at TSMC
- ·Supply tightness is tracked on /hardware
- ·TPU v7 is a chip built to run AI, made by Google
- ·It has 192 GB of fast memory so large models fit
- ·Companies buy or rent these to train and serve AI models
Frequently Asked Questions
Read the primary sources
- Ironwood announcementcloud.google.com