NVIDIA H200 vs AMD Instinct MI300X
What each costs to rent by the hour across the providers we track — NVIDIA H200 on 59 providers, AMD Instinct MI300X on 8.
As of , the NVIDIA H200 has a median rate card of $4.56 per GPU-hour versus $2.65 for the AMD Instinct MI300X — $1.92 (42%) more for 27% less memory.
Side by side
| NVIDIA H200 | AMD Instinct MI300X | |
|---|---|---|
| Architecture | Hopper | CDNA 3 |
| VRAM | 141 GB | 192 GB |
| Form factor | SXM | OAM |
| Median rate card, per GPU-hour | $4.56 | $2.65 |
| Cheapest listed rate | $1.60 Yotta Labs, spot captured | $1.11 Microsoft Azure, spot captured |
| Median per GB of VRAM, per hour | $0.032 | $0.014 |
| Providers listing it | 59 | 8 |
Providers that list both
| Provider | NVIDIA H200 | AMD Instinct MI300X | MI300X vs H200 |
|---|---|---|---|
 Seeweb | $2.93 captured | $2.65 captured | −$0.28 (−10%) |
 Paperspace | $4.47 GPU Droplets captured | $2.59 GPU Droplets captured | −$1.88 (−42%) |
 RunPod | $5.29 Secure Cloud captured | $2.39 Secure Cloud captured | −$2.90 (−55%) |
 Oracle Cloud (OCI) | $10.00 captured | $6.00 captured | −$4.00 (−40%) |
 Microsoft Azure | $10.60 westus2 captured | $6.00 eastus2 captured | −$4.60 (−43%) |
Medians are median rate cards — each provider once, at its cheapest on-demand rate, community capacity and marketplace floors out — the same figure as on each card's page; the cheapest listed rate may be spot or a community tier and is labelled as such. Compare providers directly on the overview, or see each GPU's full table: NVIDIA H200, AMD Instinct MI300X. On your own page: embed this comparison (/embed/gpus/h200-vs-mi300x, an iframe, no script).
Data updated: