neocloud · bare metal · Layer 2.5
Fluidstack
Frontier labs and very large buyers contracting dedicated campuses. It is not a place to rent one GPU.
high portability Portable. The workload is standard enough to move to another vendor without a rewrite.
What you are buying
- Business model
- Purpose-built GPU cloud. Owns or leases its own datacentre capacity and sells it directly.
- How you reach it
- You get the physical machine. Maximum control, and every layer above it is yours to run.
- Accelerators
- Not published. Its own site describes building datacentres, not a chip catalogue, and Anthropic's announcement names no part.
- Regions
- Texas and New York for the Anthropic build (Anthropic, 12 Nov 2025). No other location is published on its own site.
- Pricing model
- Not published. fluidstack.io/pricing returned a 404 on 17 Nov 2026, and the site carries no rate card.
- Getting started
- No self-serve route. The home page is a mission and hiring page with no product to buy and no sign-up link (read 17 Nov 2026).
- Capacity
- Campus scale. Its site lists building 250MW datacentres in six months and a goal of hundreds of gigawatts over the next five years; Anthropic said its Fluidstack sites come online through 2026.
- Ownership
- Private; chief executive Gary Wu (named in Anthropic's announcement). An $830M Series A at a $7.5B valuation, led by Situational Awareness (SiliconANGLE, 21 Jul 2026), then a $1.5B round led by Jane Street at an $18B valuation (Dealroom, 28 Sep 2026, citing 36Kr; Dealroom gives the earlier round as $750M, so the two accounts differ).
Verified 2026-11-17. Fields reading “not published” are exactly that — we do not estimate a figure a vendor withholds.
The counterparty
A frontier-lab builder. Anthropic announced a $50B programme of US datacentres built with Fluidstack in Texas and New York, and SiliconANGLE's July report names Anthropic as its only named customer; Dealroom adds Meta, Mistral and Poolside, and describes about $2.6B of equity beside tens of billions in debt for construction. The risk sits in building on time and in a handful of very large customers, not in an hourly rate.
A multi-year GPU commitment is a credit decision wearing a cloud contract. This is the section no benchmark covers and the one that decides what happens to your workload in 2028.
The catch
There is nothing to sign up for. The pricing page is gone (a 404 on 17 Nov 2026) and the site is about building datacentres, so for anything short of a dedicated campus it is not a supplier today.
The economics
The business is a few very large, long contracts, with construction financed largely by debt on Dealroom's account. For a buyer of that size the questions are the build schedule and who else shares the campus, not the rate.
No rate is quoted on this page on purpose. Published list prices at this layer move weekly and essentially nobody signing a real contract pays them. Dated, sourced figures live in the GPU cost calculator, where the cheapest and dearest published on-demand price for the identical H100 sit 5.6× apart.
The layers underneath this one
Whatever you rent here is a chip in a building that needs power. In 2026 megawatts, not silicon, are the binding constraint on the whole industry — a frontier rack draws 120–200 kW against a 2026 average near 27 kW, and the US interconnection queue exceeds 2,600 GW.
Layer 2 — Silicon · Layer 1 — Energy · The interconnection queue · Tokens per watt
Related providers
- OVHcloud — neocloud, high portability
- TensorWave — neocloud, high portability
- Voltage Park — neocloud, high portability
- Civo — neocloud, high portability
- CoreWeave — neocloud, high portability
- GMI Cloud — neocloud, high portability