NVIDIA H100
Hopper SXMPCIe Recent
80–94 GB VRAM · 2,000–3,900 GB/s memory bandwidth · On-demand from $0.000 /GPU/h at Scaleway
On-demand per interface: SXM from $0.000 · PCIe from $0.000 /GPU·h
Tech specs
Cloud rental prices
29 providers · On-demand $0.000 – $15.98 / Reserved $1.75 – $5.75 / Spot $0.79 – $6.30 /GPU/h
Prices checked 54m ago. Listed prices for new instances; volume discounts may differ.
| Provider | On Demand /GPU/h | Reserved /GPU/h | Spot /GPU/h |
|---|---|---|---|
| Scaleway | $0.000 | — | — |
| Jarvislabs | $0.070 | — | — |
| Latitude.sh | $1.68 | — | — |
| Voltage Park | $1.99 | — | — |
| RunPod | $1.99 | — | — |
| LeaderGPU | $2.24 | — | — |
| Denvr | $2.45 | — | — |
| Hyperstack | $2.50 | $1.75 | $2.00 |
| GMI Cloud | $2.60 | — | — |
| AceCloud | $2.64 | — | — |
| Massed Compute | $2.73 | — | — |
| Sesterce | $2.75 | — | — |
| CloudRift | $3.00 | $2.55 | — |
| OVHcloud | $3.18 | — | — |
| Thunder Compute | $3.20 | — | — |
| Leafcloud | $3.46 | — | — |
| Lambda | $3.79 | — | — |
| Together AI | $3.99 | — | $1.99 |
| Cirrascale | $4.28 | — | — |
| Neysa | $4.39 | $2.13 | — |
| DigitalOcean | $4.41 | $3.26 | — |
| Nebius | $4.50 | — | $0.79 |
| Verda | $5.40 | — | $2.70 |
| Crusoe | $5.50 | — | $6.30 |
| CoreWeave | $6.16 | — | $2.44 |
| Shakti Cloud | $6.26 | $4.13 | — |
| AWS | $6.88 | — | — |
| Azure | $6.98 | — | $1.29 |
| Oracle Cloud | $10.00 | — | — |
Similar GPUs
FAQ
How much does a NVIDIA H100 cost to rent?
On-demand NVIDIA H100 rental starts at $0.000 /GPU/h at Scaleway among 29 providers tracked (SXM from $0.000, PCIe from $0.000) — prices checked 54m ago.
How much VRAM does the NVIDIA H100 have?
The NVIDIA H100 ships with 80–94 GB of VRAM and 2,000–3,900 GB/s memory bandwidth.
Which cloud providers offer the NVIDIA H100?
29 providers list NVIDIA H100 capacity, including Scaleway, Jarvislabs, Latitude.sh. The table above compares live prices.
Specifications from official manufacturer documents. Prices as reported by providers. Non-USD prices converted at the exchange rates published in instances.json (fx_rates). Report errors: feedback [at] cheapergpus [dot] com