Google TPU Ironwood
v7 TPU pods scaling to 9,216 chips, available through Google Cloud and used for Gemini.
Full Google TPU Ironwood record →Head to head
Google TPU Ironwood vs AWS Trainium2
Google TPU Ironwood (Google) and AWS Trainium2 (AWS) both sit in Chips & Compute. Google TPU Ironwood shipped more tracked changes in the last 30 days (11 vs 7). Every value below comes from the latest crawl of the vendors' own pages.
| Attribute | Google TPU Ironwood | AWS Trainium2 AWS |
|---|---|---|
| Segment | Chips & Compute | Chips & Compute |
| Version | 7th-generation (Ironwood) | Trainium2 |
| Entry cost | from $5.4/unit | Pricing is not published on the page. Available via EC2 instances/UltraServers. |
| Pricing tiers | On Demand (us-central1) $12 · DWS Flex-start (us-central1) $6 · DWS Calendar Mode (us-central1) $8.4 · 1-year Commitment (us-central1) $8.4 · 3-year Commitment (us-central1) $5.4 · On Demand (europe-west2) $13.2 · DWS Flex-start (europe-west2) $6 · DWS Calendar Mode (europe-west2) $8.4 | — |
| Model stack | Gemini · Google TPU Ironwood · Ironwood · Ironwood (TPU v7) · Ironwood TPU · Ironwood TPU (7th Generation) · TPU v7 · TPU7x · TPU7x (Ironwood) · XLA compiler | AWS Trainium2 · Llama · Llama 3.1 · Llama 3.1 405B · Llama 3.3 · Llama 3.3 70B · Neuron compiler · Qwen3 · Stable Diffusion · Trainium2 |
| Context window | — | — |
| Public API | ||
| Routes models | ||
| Changes / 30d | 11 | 7 |
| Origin | Global | Global |
| Capabilities | 9,216-chip pods, inference optimised, liquid cooled, JAX + PyTorch/XLA, Large-scale training, Reasoning and inference, 9,216 liquid-cooled chips per pod, 42.5 ExaFlops performance | Trn2 UltraServer, 64-chip NeuronLink domain, Neuron SDK, Rainier cluster, Generative AI training, Generative AI inference, Up to 20.8 FP8 petaflops, 1.5 TB HBM3 memory |
AWS Trainium2
Trn2 UltraServers and Project Rainier capacity, programmed through the Neuron SDK.
Full AWS Trainium2 record →