AWS Trainium, Inferentia and Graviton, compared
8 Amazon Web Services chip generations, 7 deployed and 1 announced, sampling or in preview. Top stated FP8: AWS Trainium2 at 1,299 TFLOPS per chip. Largest fixed memory entry: AWS Trainium3 with 144 GB. Largest pod: AWS Trainium3 at 144 chips.
| AWS Graviton5 | Amazon Web Services | In-house (Annapurna Labs) | TSMC 3 nm | — | — | — | — | — | — | — | — | — | — | — | Deployed | 2,0251 | 2026-11-11 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AWS Trainium4 | Amazon Web Services | — | — | — | — | — | — | — | — | — | — | — | — | — | Announced | 2,0251 | 2026-11-11 |
| AWS Trainium3 | Amazon Web Services | Alchip; Marvell | 3 nm | 1443 | 4.93 | — | 2,5171 | — | 6711 | — | — | 2,5171 | — | 1442 | Deployed | 2,0245 | 2026-11-11 |
| AWS Graviton4 | Amazon Web Services | In-house (Annapurna Labs) | — | — | — | — | — | — | — | — | — | — | — | — | Deployed | 2,0231 | 2026-11-11 |
| AWS Trainium2 | Amazon Web Services | In-house (Annapurna Labs) | — | 961 | 2.91 | 1,2991 | — | — | 6671 | — | — | — | — | 642 | Deployed | 2,0234 | 2026-11-11 |
| AWS Inferentia2 | Amazon Web Services | In-house (Annapurna Labs) | — | 321 | 0.881 | — | — | — | 1901 | — | — | — | — | — | Deployed | 2,0223 | 2026-11-11 |
| AWS Trainium | Amazon Web Services | In-house (Annapurna Labs) | — | 321 | 0.821 | — | — | — | 1901 | — | — | — | — | 161 | Deployed | 2,0203 | 2026-11-11 |
| AWS Inferentia | Amazon Web Services | In-house (Annapurna Labs) | — | 81 | 0.051 | — | — | — | 641 | — | — | — | — | 162 | Deployed | 2,0183 | 2026-11-11 |
Every chip in this list
All 42 chips- AWS Graviton5192 cores · TSMC 3 nm · Deployed
- AWS Trainium4Announced
- AWS Trainium3144 GB HBM3E · 2,517 MXFP8 TFLOPS · Deployed
- AWS Graviton496 cores · Deployed
- AWS Trainium296 GB HBM · 1,299 FP8 TFLOPS · Deployed
- AWS Inferentia232 GB HBM · 190 BF16 TFLOPS · Deployed
- AWS Trainium32 GB HBM · 190 BF16 TFLOPS · Deployed
- AWS Inferentia8 GB DRAM (not HBM) · 64 BF16 TFLOPS · Deployed