The AMD MI355X (CDNA 4, 288 GB HBM3e) rents from $3.56/hr, averaged across 9 cloud providers, updated 2026-11-15. On-demand pricing has fallen 4% over the past 90 days.
AMD's Blackwell-class accelerator — 288 GB HBM3e, the most memory of any mainstream part, for the largest inference workloads. CDNA 4 · 288 GB HBM3e · 8 TB/s memory bandwidth · launched 2025.
▮ on-demand /hr · ▮ spot /hr
| Date | On-demand | Spot |
|---|---|---|
| 2026-11-15 | $8.25 | $3.56 |
| 2026-11-14 | $8.04 | $3.44 |
| 2026-11-13 | $8.04 | $3.44 |
| 2026-11-12 | $8.04 | $3.44 |
| 2026-11-11 | $8.04 | $3.44 |
| 2026-11-10 | $8.04 | $3.44 |
| 2026-11-09 | $8.04 | $3.44 |
| 2026-11-08 | $5.99 | $3.44 |
| 2026-11-07 | $5.99 | $3.44 |
| 2026-11-06 | $5.99 | $3.44 |
| 2026-11-05 | $5.99 | $3.44 |
| 2026-11-04 | $5.99 | $3.44 |
| 2026-11-03 | $5.99 | $3.44 |
| 2026-11-02 | $5.99 | $3.44 |
| GPU | On-demand /hr | Spot /hr | Memory | |
|---|---|---|---|---|
| MI355X | $8.25 | $3.56 | 288 GB HBM3e | |
| B200 | $8.28 | $4.80 | 192 GB HBM3e | MI355X vs B200 → |
| H200 | $4.59 | $3.36 | 141 GB HBM3e | MI355X vs H200 → |
| H100 | $4.01 | $2.40 | 80 GB HBM3 | MI355X vs H100 → |
| MI325X | $3.11 | $— | 256 GB HBM3e | MI355X vs MI325X → |
| MI300X | $2.86 | $1.84 | 192 GB HBM3 | MI355X vs MI300X → |
The MI355X averages $8.25/hr on-demand, $3.56/hr spot, and $4.29/hr reserved across 9 cloud providers (updated 2026-11-15).
The lowest tier is spot at about $3.56/hr on average; reserved (1–3 year commitment) runs ~$4.29/hr. Compare live per-provider rates on the Bargo GPU price tracker.
At on-demand $8.25/hr, one MI355X runs about $6,023/month (730 hours); reserved pricing lowers that to roughly $3,132/month.
On-demand, the MI355X is $8.25/hr vs the H100 at $4.01/hr — pricier per hour. See the full MI355X vs H100 comparison for specs and cost-per-job context.
Spot ($3.56/hr) is interruptible spare capacity at a discount to on-demand ($8.25/hr); use spot for checkpointed or fault-tolerant jobs and on-demand or reserved for long, uninterrupted runs.
Cross-provider averages from public cloud GPU listings. Informational only.