Hyperstack · Hourly
H100 SXM 80 GB
Compare H100 SXM 80 GB rental prices · Hyperstack pricing and terms
H100 SXM 80 GB
USD 2.72 / hr per GPU · Starting price
Stale source check: older than 72 hours, missing or future-dated. Recheck the vendor before relying on the price.
- Instance total
- Minimum instance size is unverified; the price is per GPU only.
- Configuration
- Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed.
- Minimum commitment
- Not verified.
- Availability / region
- Stock not verified · North America and Europe
- Limitations
- Published reservation starting rate; term and capacity require confirmation.
- Source checked
Prices and availability can change. Confirm the current offer with the provider.
Host resources & extra charges
Resources are per instance. Host RAM is separate from GPU VRAM; GB and GiB retain the source's units.
- Country
- Unspecified / multiple countries
- Provider zone
- Unknown
- Source location / coverage
- North America and Europe
- Cloud tier
- Unknown
- Host CPU
- Unknown
- Host RAM
- Unknown
- Included local storage
- Unknown · GPU flavor CPU/RAM, root disk and ephemeral local disk are included; capacity depends on the exact flavor and region. Ephemeral data is lost on hibernation/deletion.
- Boot storage
- Unknown · Included root disk capacity varies by flavor/region; hibernated root data is billed separately on Cloud-SSD storage.
- Billing minimum / rounding
- Unknown
- Additional persistent storage
- Unknown · Shared Storage Volumes / Cloud-SSD reference price is USD 0.000096774 per GB/hour; snapshots and hibernated root data also incur storage charges. Numeric estimate remains unknown rather than rounding this nine-decimal rate to fit the current schema.
- Outbound transfer
- Included / no charge · Public pricing states no ingress/egress or bandwidth fee.
- Setup fee
- Unknown
- Public IP / other required extras
- Public pricing lists a separately billed IP at 0.00672043 per hour; current documentation also shows USD 0.00672/hour. Exact order rate and IP quantity are not included in the estimate. · Public IP, extra persistent storage, snapshots and retained hibernation data are additional. Setup, applicable taxes and optional services remain unpriced.
- Taxes
- Unknown; excluded from estimate
Specifications checked September 27, 2026, 01:03 EDT. Recheck with the provider before relying on these details.
docs.hyperstack.cloud: pricebook ↗docs.hyperstack.cloud: flavors ↗docs.hyperstack.cloud: states and billing ↗
Public pricing shows maxima of 24 physical CPU cores and 240 GB RAM per GPU, not a fixed instance allocation. Exact flavor/region and availability require confirmation. Reservation is a starting rate; contract term and whole commitment require confirmation. Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote.
Source observations
Price history
Latest 90 checks. USD per GPU / hr. No estimated or backfilled history.
| Checked | Per GPU | Billing / qualification | Terms at check | Evidence |
|---|---|---|---|---|
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · unverified | Current documented NVIDIA H100 SXM flavor is eight GPUs; a purchasable one-GPU instance is not confirmed. Published rate is per GPU, so the legacy instance-size placeholder must not be used for a whole-instance quote. Published per-GPU reference rate. CPU, RAM, included root disk and local scratch allocation depend on the selected flavor and region; pricing-page maxima are not guaranteed instance resources. Shared Storage Volumes and public IP are separately billed. | Historical capture unavailablePricing evidence ↗ | |
| USD 2.72 | Hourlyfrom · | Historical capture unavailablePricing evidence ↗ |