Pyyan / Compare / MI400 vs B300 Blackwell Ultra vs TPU v7 Ironwood
GPU & AI Chips · verified 13 Aug 2026
| Specification | MI400 | B300 Blackwell Ultra | TPU v7 Ironwood |
|---|---|---|---|
| Summary | Double NVIDIA's memory capacity. | 15 PFLOPS dense FP4. | Inference-oriented, Google Cloud only. |
| Memory | 432GB HBM4 | 288GB HBM3e | 192GB HBM3e |
| Bandwidth | 19.6 TB/s | 8 TB/s | 7.37 TB/s |
| Price | — | ~$40–50K (verify) | — |
| Vendor | AMD | NVIDIA | |
| Category | GPU & AI Chips | GPU & AI Chips | GPU & AI Chips |
| Official | AMD ↗ | NVIDIA ↗ | Google ↗ |
Highlighted rows are where these differ.
Best for large-model inference.
Best for frontier training.
Best for GCP inference.