Best Neocloud / Neocloud companies
Neocloud companies
The neocloud providers buyers shortlist in 2026, where they're based, what they're good for, and which have Blackwell Ultra.
The biggest neocloud companies in 2026 are CoreWeave, Nebius, Lambda and Crusoe. Behind them sit dozens of smaller GPU clouds: SemiAnalysis rated 84 of them in ClusterMAX 2.0 (source). Here are the ones buyers ask us about most.
List of neocloud companies
Ordered roughly by how often they come up in large GPU buys. "Blackwell Ultra" shows whether we've confirmed a public B300 or GB300 offer; see the B300 directory for sources and dates.
| Company | Based | Known for | Blackwell Ultra |
|---|---|---|---|
| CoreWeave | US | Largest neocloud by revenue; reserved clusters for AI labs and big tech | HGX B300 GA |
| Nebius | Netherlands | Mostly owned data centers in Europe and the US; self-service and reserved | HGX B300 |
| Lambda | US | On-demand instances, 1-Click Clusters and single-tenant superclusters | GB300 NVL72, HGX B300 |
| Crusoe | US | Builds its own AI data center campuses; cloud sold through sales | GB300 NVL72, HGX B300 |
| Together AI | US | GPU clusters plus a large managed inference platform | Researching |
| Fluidstack | UK | Large dedicated clusters for AI labs | Researching |
| Voltage Park | US | Bare-metal NVIDIA clusters | Researching |
| Nscale | UK | European AI cloud and data center developer | Researching |
| Firmus | Australia | AI cloud in Asia-Pacific; ClusterMAX 3.0 Silver | Researching |
| TensorWave | US | AMD Instinct GPU cloud; ClusterMAX 3.0 Silver | AMD, not NVIDIA |
| GMI Cloud | US | NVIDIA GPU cloud with Asia-Pacific capacity; ClusterMAX 3.0 Silver | Researching |
| RunPod | US | Self-service GPU pods and serverless GPUs for developers | Researching |
| Vast.ai | US | Marketplace renting GPUs from many independent hosts | Marketplace |
| Vultr | US | General cloud with a large GPU offer | Researching |
Not on the list: AWS, Microsoft Azure, Google Cloud and Oracle. They sell plenty of GPUs, but they're hyperscalers, not neoclouds.
How the neocloud market splits
The large reserved-cluster players
CoreWeave and Nebius sell thousands of GPUs at a time to AI labs and big tech, on multi-year contracts. Both were rated Platinum, the top tier, in SemiAnalysis ClusterMAX 3.0 in September 2026 (summary). Read CoreWeave vs Nebius for the differences.
Builders and operators
Crusoe, Nscale and Fluidstack are as much data center developers as clouds. They're often the route to very large, single-tenant builds, but the capacity may still be under construction.
Developer-first GPU clouds
Lambda, Together AI and RunPod make it easy to start with a few GPUs and grow. Lambda also sells large GB300 NVL72 superclusters.
Marketplaces
Vast.ai and similar marketplaces rent GPUs from many independent hosts. They're good for cheap experiments, less so for a connected cluster with one accountable operator.
Which neocloud should you use?
It depends on size and timing more than brand:
- 1 to 8 GPUs, today: a self-service neocloud such as Lambda, Nebius or RunPod.
- 16 to 256 GPUs for a few months: instant or reserved clusters from Lambda, Nebius, Together AI or CoreWeave.
- 256+ GPUs, or GB300 NVL72 racks: get quotes from several of CoreWeave, Nebius, Lambda and Crusoe. Capacity at this size moves week to week. We can ask them for you.
Questions
What are the biggest neocloud companies?
How many neoclouds are there?
Are neocloud providers safe for long contracts?
Is Lambda a neocloud?
RackShore runs Best Neocloud and may be paid a referral fee by a provider. That doesn't change which providers are listed or how they're described. Spot a mistake? Tell us.