Fluidstack has reportedly closed a US$1.5 billion funding round led by Jane Street as the AI infrastructure company shifts from finding spare GPUs to building and operating vast compute facilities. The financing would more than double its valuation from US$7.5 billion earlier this year, according to Forbes reporting cited by VnExpress International. That makes Fluidstack’s expansion worth watching for a reason beyond the funding headline: the AI race is increasingly being fought by the companies supplying the electricity, chips and buildings underneath the models.
Watch Desk analysis
What happened
Fluidstack began in 2017 as a marketplace connecting owners of idle gaming GPUs with AI researchers who needed computing power. It later moved into renting larger GPU pools and now builds infrastructure for customers including Meta, Mistral, Poolside and Black Forest Labs.
Its biggest reported project is with Anthropic. The Claude developer selected Fluidstack to build custom data centres in New York and Texas as part of a planned US$50 billion investment in American computing infrastructure. Fluidstack is also involved in Anthropic’s planned deployment of up to one million Google Tensor Processing Units, or TPUs, specialised chips designed for AI workloads.
The funding and project details remain reported rather than confirmed by a company announcement in the evidence supplied. The French data-centre project announced in 2025 also fell away after Fluidstack shifted focus towards larger US contracts, so planned capacity is not the same thing as delivered capacity.
The numbers
- The reported raise
A closed US$1.5 billion Jane Street round would give a specialist AI infrastructure operator unusually deep financing. - The business has changed shape
Fluidstack has moved from brokering unused GPUs to operating large-scale infrastructure for major AI customers. - A planned TPU deployment
The Anthropic project could make Fluidstack the first publicly known TPU data-centre operator outside Google, according to Forbes. - The power target is enormous
An investor report projects 1.3 gigawatts across more than 10 sites this year, while the company says it aims for 50 gigawatts by 2030. - The projections need concrete proof
Sites, grid connections, operating capacity and customer contracts will matter more than another impressive spreadsheet.
Why it matters
AI companies cannot scale on model cleverness alone. They need reliable access to accelerators, power and facilities, which gives infrastructure operators more leverage over what gets built, where services run and how quickly new systems can expand.
It also broadens the compute supply chain beyond the familiar cloud giants. If Fluidstack and similar operators can turn financing into functioning GPU and TPU sites, AI customers may gain more routes to scarce capacity. If they cannot, the industry will have funded a very expensive collection of future tense.
Our read
The important development is the rise of the specialist AI infrastructure operator, not the founders’ billionaire maths. Treat the reported US$1.5 billion round and the power forecasts as attributed claims, then look for operational evidence before using them in procurement or investment decisions.
What to watch
- Confirmation or denial of the reported Jane Street financing and revised valuation.
- Anthropic’s New York and Texas facilities becoming operational.
- Evidence that planned TPU deployments are delivered and running.
- Secured sites, grid connections and customer contracts behind the 1.3GW and 50GW figures.
Discussion spark: How much of the next AI infrastructure wave should readers credit to model companies themselves, and how much to specialised operators such as Fluidstack?
Sources and evidence
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.