Impossible Cloud
Bare metal GPU clusters on committed terms, S3 storage and managed inference
Last reviewed Jun 12, 2026
Impossible Cloud is a Hamburg-based cloud provider offering S3-compatible object storage, dedicated bare metal GPU clusters and a managed inference API for open-weight models. GPU clusters (NVIDIA H100, H200, B200, B300, GB300 NVL72, L40S and RTX workstation cards) are listed with a published per-GPU-hour rate for each contract term, from one month up to five years, in EU, US and APAC regions. The managed inference API is OpenAI-compatible, billed per million tokens, and in early access.
Available GPUs
Hourly on-demand pricing. Click column headers to sort.
Prices last updated: September 24, 2026
| B200 | $8.42/hrReserved · 7moEU 8× | 8× | |||
| B200 | $4.21/hrReserved · 60moUS 8× | 8× | |||
| GB300 | $10.00/hrReservedEU 72× | 72× | |||
| H100 PCIe | $2.26/hrReserved · 6moEU 8× | 8× | |||
| H100 PCIe | $2.09/hrReserved · 12moEU 8× | 8× | |||
| H100 SXM | $2.87/hrReserved · 12moAPAC 8× | 8× | |||
| H200 | $4.00/hrReserved · 1mo 8× | 8× | |||
| H200 | $3.68/hrReserved · 3mo 8× | 8× | |||
| H200 | $3.48/hrReserved · 12mo 8× | 8× | |||
| HGX B300 | $4.32/hrReservedEU 8× | 8× | |||
Impossible Cloud pricing by GPU
Configurations, price rank, and alternatives for one GPU at a time.
Pros & Cons
Advantages
- Per-GPU-hour rates published for every listed cluster and contract term
- GPU compute and S3 object storage from one provider, so data and compute can sit in the same region
- Dedicated bare metal hardware without virtualization
- German headquarters and European data residency options
- Quotes by email without a sales call
Limitations
- No on-demand hourly rate; all GPU capacity is sold on committed terms
- Many listed clusters have a future availability date rather than immediate capacity
- Some listings show region, configuration or node count only on request
- Managed inference is in early access with pricing given on request
Key Features
Bare Metal GPU Clusters
Dedicated NVIDIA systems without virtualization, from 2-GPU PCIe servers to 8-GPU SXM nodes and GB300 NVL72 racks, with InfiniBand or RoCE interconnect on larger clusters.
Published Term Pricing
Each listed cluster shows a per-GPU-hour rate for every available contract term, from one month to 60 months, along with region, node count and availability date.
Public Pricing Feed
The GPU offer list is published as an unauthenticated JSON feed, regenerated daily.
S3-Compatible Object Storage
Object storage compatible with the AWS S3 API, with no egress or API-call fees and no minimum storage duration.
Managed Inference
OpenAI-compatible inference endpoints for open-weight models (DeepSeek, Qwen, Kimi, GLM, gpt-oss, Nemotron), billed per million tokens; currently in early access.
Compliance
Headquartered in Germany with ISO 27001 certification and GDPR-aligned data handling in European data centers.
Pricing Options
| Option | Details |
|---|---|
| Committed GPU Terms | Per-GPU-hour rates for contract terms from one month to 60 months; longer terms carry lower rates. |
| Storage Pay-as-you-go | Object storage billed per TB per month with no egress or API-call fees. |
| Storage Reserved Capacity | Discounted per-TB storage pricing for 1, 3 and 5 year capacity commitments, sold through partners. |
| Per-Token Inference | Managed inference billed per million tokens, with rates provided on request during early access. |
Availability & Support
Regions
GPU clusters listed in EU, US and APAC regions; object storage in European and US data centers
Support
Customer support portal, knowledge base and documentation; regional consultants for quotes and deployment.
Company
Published by Impossible Cloud to identify the company behind this listing.
- Legal entity
- Impossible Cloud GmbH
- Headquarters
- 🇩🇪 Hamburg, Germany
- Registration
- Handelsregister, District Court Hamburg HRB 170423
Getting Started
- 1
Browse Available Clusters
Filter the bare metal GPU listings by GPU model, region and contract term.
- 2
Request a Quote
Request a listed cluster or a custom configuration; quotes are sent by email.
- 3
Agree Term and Deployment
Confirm the contract term, region and start date for the cluster.
- 4
Manage in the Console
Use the Impossible Cloud management console for GPU servers and storage.
Frequently Asked Questions
What GPU types does Impossible Cloud offer?
Impossible Cloud offers various GPU types including H100 SXM, H200, H200, H200, HGX B300, HGX B300, HGX B300, HGX B300, HGX B300, HGX B300, HGX B300, HGX B300, HGX B300, HGX B300, GB300, L40S, L40S, L40S, H100 PCIe, H100 PCIe, B200, B200. Check the pricing table above for current availability and pricing.
How do I get started with Impossible Cloud?
Browse Available Clusters, Request a Quote, Agree Term and Deployment, Manage in the Console
What are Impossible Cloud's main advantages?
Impossible Cloud's main advantages include: Per-GPU-hour rates published for every listed cluster and contract term, GPU compute and S3 object storage from one provider, so data and compute can sit in the same region, Dedicated bare metal hardware without virtualization, German headquarters and European data residency options, Quotes by email without a sales call.
What are Impossible Cloud's limitations?
Impossible Cloud's main limitations include: No on-demand hourly rate; all GPU capacity is sold on committed terms, Many listed clusters have a future availability date rather than immediate capacity, Some listings show region, configuration or node count only on request, Managed inference is in early access with pricing given on request.
Compare Providers
Find the best prices for the same GPUs and models from other providers
