| GPUs offered | Regions | Ownership | Bare metal | Slurm | Kubernetes | ||||
|---|---|---|---|---|---|---|---|---|---|
| Voltage Park | $1.992 | H100, H200, B200, GB300 | US | privateLightning AI | Yes | — | Yes | — | 2026-11-09 |
| Denvr Dataworks | $2.451 | GB200, B200, H200, GH200, H100, A100 | Canada | private | Yes | — | — | — | 2026-11-09 |
| Core42 | $2.501 | H100, H200, B200, GB300, MI300X, MI355X | UAE, US, Europe | subsidiaryG42 | Yes | Yes | Yes | Participation | 2026-11-09 |
| GMI Cloud | $2.601 | H100, H200, B200, GB200, GB300 | N. America, Europe, APAC | private | Yes | Yes | Yes | Silver | 2026-11-09 |
| OVHcloud | $2.991 | H200, H100, A100, L40S, L4 | France, Canada | public | — | — | Yes | Underperforming | 2026-11-09 |
| Vultr | $2.991 | B200, H100, GH200, A100, L40S, MI355X, MI325X, MI300X | 34 regions worldwide | private | Yes | Yes | Yes | Participation | 2026-11-09 |
| Massed Compute | $3.141 | B300, B200, H200, H100, A100, L40S, L40, RTX PRO 6000 | — | private | Yes | — | — | Underperforming | 2026-11-09 |
| Hyperbolic | $3.192 | H100, H200, B200 | — | private | Yes | — | Yes | Participation | 2026-11-09 |
| Hyperstack | $3.201 | B300, B200, GB300, H200, H100, A100, RTX Pro 6000, L40 | Europe, N. America | privateNexGen Cloud | — | — | Yes | Bronze | 2026-11-09 |
| Crusoe | $3.901 | Vera Rubin NVL72, GB300, B300, GB200, B200, H200, H100, A100, L40S, MI355X, MI300X | US, Norway, Paraguay | private | — | Yes | Yes | Bronze | 2026-11-09 |
| RunPod | $3.991 | H100, H200, B200, B300, A100, L40S, L40, L4, RTX PRO 6000, MI300X | — | private | — | Yes | No | Participation | 2026-11-09 |
| Together AI | $3.991 | GB300, GB200, B300, B200, H200, H100 | US, Europe | private | — | Yes | Yes | Bronze | 2026-11-09 |
| DigitalOcean GPU Droplets | $4.411 | B300, H200, H100, MI350X, MI325X, MI300X, L40S | US, Canada, EU | publicDigitalOcean Holdings, Inc. | Yes | — | Yes | Bronze | 2026-11-09 |
| Nebius | $4.501 | GB300, B300, GB200, B200, H200, H100, RTX PRO 6000, L40S | EU, UK, US, Israel | public | — | — | Yes | Platinum | 2026-11-09 |
| Lambda | $4.891 | B200, H100, GH200, A100 | US, Japan, India, Germany, Israel | private | — | Yes | Yes | Silver | 2026-11-09 |
| Verda | $4.9871 | GB300, B300, B200, H200, H100, A100, L40S, RTX PRO 6000 | EU | private | — | Yes | Yes | Bronze | 2026-11-09 |
| Lightning AI | $5.451 | L4, L40S, RTX Pro 6000, A100, H100, H200, B200, GB300 | US | private | Yes | Yes | Yes | — | 2026-11-09 |
| CoreWeave | $6.161 | GB300, GB200, B300, B200, RTX PRO 6000, GH200, H200, H100, L40S, L40, A100 | US, Europe | public | Yes | Yes | Yes | Platinum | 2026-11-09 |
| ChronoScale | — | VR200 NVL72, GB300 NVL72, HGX B300, HGX H100 | US, SE Asia, Nordics | publicApplied Digital (NASDAQ: APLD), majority shareholder | Yes | — | — | — | 2026-11-09 |
| Cirrascale Cloud Services | — | GB300, B300, B200, RTX PRO 6000, MI355X | US | private | — | — | — | — | 2026-11-09 |
| CUDO Compute | — | H100, H200, B100, B200, B300, GB200, GB300, VR200 | N. America, Europe, UK, Middle East | private | Yes | — | — | — | 2026-11-09 |
| Firmus | — | GB300, H200, H100, A100, L40S | Singapore, Australia | private | Yes | — | — | Silver | 2026-11-09 |
| Fluidstack | — | — | US | private | — | — | — | — | 2026-11-09 |
| Hot Aisle | — | MI300X, MI355X | US | private | Yes | — | — | — | 2026-11-09 |
| IREN | — | H100, H200, B200, B300, GB300 | US, Canada | public | Yes | — | — | Underperforming | 2026-11-09 |
| Nscale | — | GB300 | Norway, UK, Portugal, Iceland, US | private | Yes | Yes | Yes | — | 2026-11-09 |
| Ori | — | — | UK | privateRadiant (Brookfield-backed) | Yes | Yes | Yes | Participation | 2026-11-09 |
| Scaleway | — | H100, B300, L40S, L4 | France, Poland | subsidiary | Yes | — | — | — | 2026-11-09 |
| Sesterce | — | — | France, Finland | private | Yes | — | — | — | 2026-11-09 |
| Shakti Cloud | — | H100, L40S, B200, B300 | India | privateYotta Data Services | Yes | Yes | Yes | — | 2026-11-09 |
| Taiga Cloud | — | H100, H200 | UK, Portugal, Sweden, Norway | subsidiaryNorthern Data Group, part of Rumble’s Quake AI (RUM Group, Nasdaq: RUM) | Yes | Yes | Yes | — | 2026-11-09 |
| TensorDock | — | H100, A100, L40 | 20+ countries | private | — | — | — | — | 2026-11-09 |
| TensorWave | — | MI300X, MI325X, MI355X, MI455X | N. America | private | Yes | Yes | Yes | Silver | 2026-11-09 |
| Vast.ai | — | B200, GB200, H100 | — | private | — | — | — | Participation | 2026-11-09 |
Price watch
How prices are readOwnership changes
7 on the list- Voltage Park Merged with Lightning AI, Jan 2026
- Lightning AI Merged with Voltage Park, Jan 2026
- Ori Now Radiant (Brookfield), 2026
- Taiga Cloud Now part of Rumble’s Quake AI, 2026
- Nscale Filed for NYSE IPO (NSCL), 18 Sep 2026
- Verda Renamed from DataCrunch, 2025
- DigitalOcean GPU Droplets Bought Paperspace, Jul 2023
How the index is built
Full methodologyEvery value comes from the company’s own pricing page, docs, filings or press release, or from SemiAnalysis ClusterMAX for the rating. The small number beside a figure opens its source, with the date we read it. A blank cell means the company has no public figure for it yet.
Prices are on-demand list prices for one GPU, in US dollars. Quote-only, host-set and local-currency prices stay out of the price column and are spelled out on the company’s page instead.
Spot a stale number? Email hello@neocloudcompanies.com with the source and the row is re-checked and re-dated. Featured placement is paid and labelled; every row stays exactly as sourced.
Questions buyers and investors ask
What is a neocloud?
A neocloud is a cloud company built to rent GPU compute for AI, separate from the general-purpose hyperscalers (AWS, Azure, Google Cloud). It sells NVIDIA or AMD GPUs by the hour or on multi-month to multi-year contracts, usually as InfiniBand or RoCE clusters with Kubernetes or Slurm on top. SemiAnalysis uses the term in ClusterMAX, its rating of these providers. CoreWeave, Nebius, Lambda and Crusoe are the best-known examples.
Sources SemiAnalysis ClusterMAX · ClusterMAX 2.0 (SemiAnalysis)
How much does an H100 cost per hour on a neocloud?
Published on-demand list prices for one H100 ran from $1.99 to $6.16 per GPU-hour on 9 Nov 2026. Voltage Park lists from $1.99, Denvr $2.45, GMI Cloud $2.60, OVHcloud and Vultr $2.99 (Vultr as an 8-GPU bare-metal plan), Crusoe $3.90 (rising to $4.50 on 1 Dec 2026), RunPod Secure Cloud $3.99, Nebius $4.50, Lambda $4.89 (price list effective 16 Nov 2026) and CoreWeave $6.16 (8-GPU instances only). Commit to a term and the rate drops: DigitalOcean charges $4.41 on demand and $3.26 on a 12-month contract.
Sources Voltage Park pricing · Crusoe Cloud pricing · Nebius prices · Lambda pricing · CoreWeave pricing · DigitalOcean GPU Droplets pricing · Vultr bare-metal plans (API)
Which neocloud companies are publicly listed?
CoreWeave (Nasdaq: CRWV), Nebius (Nasdaq: NBIS), IREN (Nasdaq: IREN), ChronoScale (Nasdaq: CHRN, Applied Digital’s former cloud business), DigitalOcean (NYSE: DOCN) and OVHcloud (Euronext Paris). Nscale filed for a NYSE IPO under the proposed ticker NSCL on 18 Sep 2026. Taiga Cloud’s parent Northern Data is now part of Rumble’s RUM Group (Nasdaq: RUM).
Sources CoreWeave Q2 2026 results · IREN Q1 FY27 results (SEC) · OVHcloud investor relations · Nscale files for IPO · ChronoScale fiscal Q1 2027 results
Which GPU clouds rate highest for cluster quality?
In SemiAnalysis ClusterMAX 3.0 (Sep 2026), CoreWeave and Nebius rate Platinum. Lambda, TensorWave, GMI Cloud and Firmus rate Silver. Crusoe, Together AI, DigitalOcean, Hyperstack and Verda rate Bronze. ClusterMAX tests real clusters on security, networking, reliability, orchestration and storage; the full review is on the SemiAnalysis site.
Which neoclouds offer AMD Instinct GPUs?
TensorWave and Hot Aisle run AMD only; Hot Aisle lists the MI300X from $2.99 per GPU-hour. Crusoe, Vultr, RunPod, DigitalOcean, Core42 and Cirrascale list AMD Instinct GPUs next to NVIDIA. Use the AMD Instinct filter above the table to see them all.
Sources Hot Aisle pricing · TensorWave MI325X
What are the best CoreWeave competitors?
On cluster quality, Nebius matches CoreWeave at Platinum in ClusterMAX 3.0. On price, Nebius lists an H100 at $4.50 per GPU-hour against CoreWeave’s $6.16, and Voltage Park, Denvr and GMI Cloud list under $3. For multi-gigawatt contracts, Crusoe (6 GW+ contracted), Nscale and IREN sell the same kind of large reserved clusters.
Sources SemiAnalysis ClusterMAX 3.0 cloud review · Nebius prices · Crusoe Series F release