From $7.5 billion to $18 billion in four months — more than double. The company called Fluidstack does something straightforward: it builds data centers specifically for AI companies.
The infrastructure provider behind Anthropic
When you use Claude, where does your request run? Part of the answer is Fluidstack. This data center company, spun out of Oxford University and later moved its headquarters to New York, counts Anthropic, Meta, Mistral, Poolside, and Black Forest Labs among its clients.
Late last year, Fluidstack secured a $50 billion partnership with Anthropic to build dedicated data centers in Texas and New York. This is not cloud services — it is custom infrastructure designed specifically for AI training and inference.
Why Fluidstack, not AWS
There is a key distinction worth understanding.
AWS, Google Cloud, and Azure serve a wide range of computing needs — web hosting, databases, traffic distribution. Fluidstack does only one thing: AI compute infrastructure.
From architecture design to cooling solutions to network interconnects, everything is optimized for large model training and inference. Companies like Anthropic and Meta are not using general-purpose clouds; they are using infrastructure tailored specifically for their workloads.
This is a narrow but extremely demand-driven business.
The valuation logic behind the four-month doubling
| Time | Event | Valuation |
|---|---|---|
| December 2025 | Completed approximately $700 million funding round | $7.5B |
| February 2026 | Google in talks to invest $100 million | — |
| April 2026 | Reportedly in talks for $1 billion new funding | $18B |
The lead investor in that December round was notable: Situational Awareness, the AGI-focused fund of former OpenAI researcher Leopold Aschenbrenner. Stripe founders the Collison brothers and former GitHub CEO Nat Friedman were also involved.
Four months later, the valuation more than doubled, reportedly led by Jane Street.
The logic from $7.5B to $18B is simple: demand for AI compute is still exploding, and there are only a handful of suppliers capable of providing specialized AI infrastructure. Fluidstack has already secured top-tier clients with large contracts in hand, so capital naturally follows.
This is not the first case of explosive growth
The AI data center and infrastructure sector is now a magnet for capital. Coreweave went public this year with a valuation that once exceeded $40 billion. Lambda Labs has been providing GPU compute for AI companies.
Fluidstack's differentiator: it does not resell someone else's servers. Instead, it builds its own data centers and signs long-term custom contracts with clients. Once a data center for Anthropic or Meta is built, the infrastructure provider locks in that client for years.
This kind of stickiness is common in SaaS, but even more solid in infrastructure.
Summary
The AI competition appears to be about models on the surface, but underneath it is about compute, and below that, data centers. What Fluidstack does is dull enough, but it sits at a position that cannot be bypassed — every company serious about AI must eventually solve where its compute comes from.
The valuation jump from $7.5 billion to $18 billion is not because it launched a new product. It is because the scarcity of this position is being recognized by more and more capital.
Sources: CocoLoop, AI data center startup Fluidstack in talks for $1B round at $18B valuation months after hitting $7.5B (TechCrunch)