MI300X vs H100
MI300X starts at $2.39/GPU-hour on-demand; H100 starts at $1.75/GPU-hour on-demand. Compare the exact variants, memory and published specifications below. Peak compute figures do not establish application speed.
26.8% cheaper at H100.
updated 2026-10-06. Price alone does not establish equivalent capacity or performance.
MI300X
H100
Spec sheet
| MI300X | H100 | |
|---|---|---|
| Spec variant | MI300X | H100 SXM |
| Architecture | CDNA 3 | Hopper |
| VRAM (spec variant) | 192 GB | 80 GB |
| FP16 TFLOPS (dense) | 1307 | 989 |
| Memory bandwidth | 5300 GB/s | 3350 GB/s |
| TDP | 750 W | 700 W |
| Interconnect | Infinity Fabric | NVLink 900GB/s |
| Launched | 2023 | 2022 |
What they cost to rent (2026-10-06)
| MI300X | H100 | |
|---|---|---|
| Cheapest on-demand $/hr | $2.39 | $1.75 |
| Cheapest spot/community $/hr | – | $0.94 |
| Cheapest source | RunPod | Lium |
| Providers offering it | 5 | 34 |
Lowest recorded on-demand rate per configuration; highlighted = cheaper. Spot/marketplace tiers shown on the individual pages.
FAQ
H100 is cheaper: the MI300X rents from $2.39 per GPU-hour and the H100 from $1.75 on-demand, updated 2026-10-06.
The largest recorded MI300X variant has 192 GB and the largest H100 variant has 94 GB. Other variants may have less memory. The MI300X has more.
Compare more rental options: the screener · price movers · methodology.