Fluidstack seeks $1 billion round at $18 billion valuation

Fluidstack is in talks to raise roughly $1 billion at an $18 billion valuation just months after seeking funding at about $7.5 billion, according to Bloomberg and reporting cited by TechCrunch. The London-founded company, which builds AI-focused data centers, has been one of the clearest beneficiaries of the scramble for compute capacity as model developers outgrow the standard cloud.

TechCrunch reports that the new round could be led by quantitative trading firm Jane Street, a notable pairing for a business whose product is effectively power, land, and procurement discipline wrapped around GPU clusters. Fluidstack’s momentum has been driven by a headline partnership: in November, Anthropic said it had signed a $50 billion agreement with the startup to build custom data centers in Texas and New York. Anthropic still relies heavily on AWS and Google Cloud for Claude, but the deal signals a desire to secure dedicated capacity and reduce dependency on general-purpose hyperscalers.

The financing story also highlights how quickly the AI infrastructure market is re-drawing the map. Fluidstack, spun out of Oxford and previously better known in Europe for work with companies such as Mistral, has relocated its headquarters from the UK to New York, TechCrunch notes. It also pulled out of a €10 billion French AI project to focus on US opportunities, according to Bloomberg. In a sector where grid interconnect queues, permitting timelines, and equipment lead times set the pace, moving closer to the deepest capital markets and the biggest buyers can matter as much as any technical differentiator.

The company’s pitch is specialization. Hyperscalers sell broad menus—storage, databases, general compute—optimized for many workloads. Fluidstack is selling facilities tuned for one workload: training and serving large models. That is attractive to customers whose compute bills are no longer a line item but the business itself. It also shifts risk: when a model company signs a bespoke buildout, it is committing to a long-lived physical footprint, while the data-center operator is betting that demand for high-density AI capacity will stay ahead of supply.

Fluidstack’s reported customer list—Anthropic, Meta, Poolside, Black Forest Labs and others—reads like a directory of companies competing for the same scarce inputs. The market is rewarding the intermediaries who can line up sites, power, and hardware fast enough to keep those model roadmaps on schedule.

The valuation jump, from roughly $7.5 billion to $18 billion on the back of talks rather than a publicly announced close, is itself a data point: in the current AI boom, the bottleneck is no longer ideas but megawatts and delivery dates.