Home / Directory / AI infrastructure / GPU & AI clouds / Fluidstack
Fluidstack
GPU neocloud and AI infrastructure operator that reserves large NVIDIA clusters for training and inference and builds the data centers that house them, including custom sites for Anthropic in Texas and New York. Forbes reported an $18 billion valuation after a September 2026 round.
AI labs and platform teams reserving hundreds to thousands of GPUs for training or heavy inference with committed terms.
Developers who only want self-serve single-GPU pods with a public hourly card, or colo buyers seeking facility leases.
Verdict
Fluidstack fits when the job is reserved GPU capacity at cluster scale and you are willing to negotiate term, SKU mix, and support. Fluidstack designs, builds and operates the data centers behind its large reserved clusters, and it does not sell retail single-card rentals. Public dollar tables are thin, so compare written quotes against CoreWeave and Nebius on-demand cards before you sign. Skip it for weekend experiments on one GPU.
Score Breakdown
How Fluidstack scores in the categories that matter to its buyers.
The Field at a Glance
Where Fluidstack sits among GPU & AI clouds vendors we reviewed, by Overall Score and relative typical engagement cost.
Fluidstack is strongest for AI labs and platform teams reserving hundreds to thousands of GPUs for training or heavy inference with committed terms. It scores 7.4, behind Nebius (8.3), CoreWeave (8.1) and Shadeform (8.1) and ahead of CUDO Compute (6.8). Its relative typical engagement cost sits mid-pack among these peers. Nebius is the stronger fit for teams that need NVIDIA GPU VMs or clusters with a published price card and managed Kubernetes or Slurm for AI.
Use-case matrix
| Use case | Fit | Notes |
|---|---|---|
| Reserved multi-hundred GPU clusters | Strong | Large reserved clusters are the core business. |
| Training capacity on a deadline | Strong | Fluidstack sells reserved capacity for large training runs. |
| Public self-serve single GPU | Poor | Single-GPU self-serve is not the main product. |
| Facility colo / MW lease | Poor | Fluidstack sells compute and does not lease colo space. |
| Inference token API | Mixed | You run your own serving stack on reserved GPUs. |
| Spot hobby workloads | Weak | Reserved, sales-led deals do not suit spot workloads. |
Review Excerpts
Excerpts from company materials and public coverage of Fluidstack, plus representative buyer views drawn from public information. Paid reviews and pay-for-play sites were excluded.
“Reserved Clusters cover 256 to 10k+ GPUs on terms of 30 days or more with monthly or annual billing and discounted rates.”
“We needed a reserved cluster for training and not one small machine at a time, and Fluidstack could deliver that scale.”
Representative buyer view
“Without a public rate card, finance had trouble benchmarking the quote. We asked for SKU, interconnect and exit terms in writing before comparing it with CoreWeave or Nebius list prices.”
Representative buyer view
“Lead times depend on the wider GPU supply market, and a reserved quote can slip when upstream capacity is tight.”
Representative buyer view
Where customer reviews couldn't be found, our analysts attempted to give a voice to buyer sentiment, drawing on real information from review sites, forums, and product documentation.
Based on: Fluidstack’s announcement of the Anthropic data centers and Anthropic’s infrastructure announcement
Company Facts
- Founded
- 2017 (CB Insights)
- Headquarters
- London, England, United Kingdom (CB Insights)
- Total funding raised
- $2.33B (CB Insights)
- Key investors
- United States Department of War, Google, Jane Street Group, Situational Awareness, BlackRock (CB Insights)
- Website
- fluidstack.io
Methodology
This page is an independent evaluation of Fluidstack for buyers comparing options in GPU & AI clouds. AI Industry Reviews accepts no sponsorships, advertising, or pay-for-placement fees. Fluidstack did not pay for this review.
What we scored
The headline number is an Overall Score on a 0-10 scale. Eight criteria fall under it in two groups.
- Buyer outcomes (for GPU & AI clouds)
- Reserved GPU capacity access
- Cluster build and delivery flexibility
- Ops support for large jobs
- Contract clarity
- Company & commercial
- Innovation & product leadership
- Project management & communication
- Pricing
- Contract fairness
Pricing measures whether the price looks fair for the value delivered, including packaging and renewal friction that show up in real buying cycles.
Score Composition
| Input | Weight | What it covers |
|---|---|---|
| Reviews | 40% | A proprietary read of what practitioners say about likes, complaints, and day-to-day use, including public review sites, forums, and private chat rooms. Paid reviews and pay-for-play sites are out of scope. |
| Product | 35% | Hands-on look at screens and workflows. |
| Pricing | 15% | Whether the price looks fair for what you get. |
| Docs & training | 10% | Docs, tutorials, and training material. |
How we balanced the evidence
The Overall Score is the simple average of the eight criteria. Recommendation language follows that score and the use-case matrix above.
Scope
Fluidstack is graded here as GPU & AI clouds. Criteria scores can move as more review volume and product checks are added.