H200 vs MI300X
Datasheet specs side by side, plus what each card actually costs to rent today across every provider we track.
H200
MI300X
Spec sheet
| H200 | MI300X | |
|---|---|---|
| Architecture | Hopper | CDNA 3 |
| VRAM (max variant) | 143 GB | 192 GB |
| FP16 TFLOPS (dense) | 989 | 1307 |
| Memory bandwidth | 4800 GB/s | 5300 GB/s |
| TDP | 700 W | 750 W |
| Interconnect | NVLink 900GB/s | Infinity Fabric |
| Launched | 2023 | 2023 |
Computed price-performance at today's cheapest on-demand rates: H200 $3.66 per PFLOP-hour vs MI300X $1.68. Formula: cheapest $/hr ÷ dense FP16 PFLOPS (1 PFLOP = 1000 TFLOPS). Specs from vendor datasheets (source, source).
What they cost to rent (2026-07-18)
| H200 | MI300X | |
|---|---|---|
| Cheapest on-demand $/hr | $3.62 | $2.19 |
| Cheapest spot/community $/hr | $3.31 | – |
| Cheapest source | Massed Compute | RunPod |
| Providers offering it | 15 | 2 |
Cheapest verified on-demand rate per side from today's snapshot; highlighted = cheaper. Spot/marketplace tiers shown on the individual pages.
FAQ
MI300X is cheaper: the H200 rents from $3.62 per GPU-hour and the MI300X from $2.19 on-demand, verified 2026-07-18.
The H200 has 143 GB and the MI300X has 192 GB. The MI300X has more.
Prices render from the live dataset each build. More: the screener · price movers · methodology.