Investors appear ready to pour yet another mountain of cash into artificial-intelligence infrastructure. According to Bloomberg, London-born data-center specialist Fluidstack is negotiating a new funding round of approximately $1 billion that could lift its valuation to roughly $18 billion. If the deal closes, the price tag would be more than twice the startup’s reported worth just a few months ago, underscoring the market’s scramble for AI-focused computing capacity.
A rapid climb in valuation
The prospective round is said to be led by quantitative-trading giant Jane Street, people familiar with the talks told Bloomberg. The numbers mark a dramatic escalation from December, when Fluidstack was reportedly seeking $700 million at a $7.5 billion valuation. That earlier raise—never formally announced—was expected to be led by Situational Awareness, an AGI-oriented fund created by former OpenAI researcher Leopold Aschenbrenner and backed by notables including Stripe co-founders Patrick and John Collison, former GitHub CEO Nat Friedman, and longtime AI investor Daniel Gross.
While the December financing round has yet to be confirmed publicly, talk of additional capital continued into February. The Wall Street Journal reported that Google was considering contributing $100 million to the same raise, illustrating the intense interest major tech companies have in securing next-generation compute.
Side-by-side comparison: December vs. spring negotiations
- December 2025: Seeking $700 million • Target valuation: $7.5 billion • Lead: Situational Awareness
- April 2026: Negotiating $1 billion • Target valuation: $18 billion • Lead: Jane Street (reported)
The implied 140 percent jump in price in less than half a year outpaces many of the high-velocity valuation hikes witnessed during the recent generative-AI boom.
Why the gold rush? One word: Anthropic
The surge in enthusiasm can be traced to a headline-grabbing partnership late last year. In November, AI safety startup Anthropic revealed a $50 billion agreement with Fluidstack to build custom data centers in Texas and New York. Rather than using general-purpose “hyperscaler” clouds like AWS or Google Cloud, Anthropic wants facilities engineered from the ground up for AI training and inference.
That contract—massive by any industry standard—delivered two immediate benefits to Fluidstack:
- Validation: Landing a top-tier AI player demonstrated that its purpose-built model can win against much larger incumbents.
- Pipeline: The Anthropic deal alone implies years of construction, hardware procurement, and managed-services revenue.
Anthropic already relies heavily on AWS and maintains partnerships with both Google and Microsoft to make its Claude models available on their platforms. Even so, surging demand for compute-intensive large language models has forced the company to seek greater control over dedicated capacity—which Fluidstack promises to deliver. The arrangement mirrors rival OpenAI’s rapid push to secure exclusive infrastructure inside Microsoft’s Azure cloud, illustrating a wider industry trend.
From Oxford to Manhattan
Fluidstack began as an academic spin-out of Oxford University before gaining traction among European AI developers. After clinching the Anthropic deal, the company shifted its headquarters from the United Kingdom to New York City, signaling a strategic pivot to win U.S. contracts and capital. In March, Bloomberg reported that the startup withdrew from a €10 billion AI project in France to concentrate resources stateside.
Expanding customer roster
Beyond Anthropic, Fluidstack lists a growing lineup of high-profile users:
- Meta—parent company of Facebook and Instagram
- Poolside—an up-and-coming generative-AI research lab
- Black Forest Labs—a stealth-mode AI venture
- Mistral—the European LLM creator that first put Fluidstack on the map
Serving these clients has allowed Fluidstack to specialize its facilities for the dense clusters of GPUs and proprietary networking architectures required by modern model training. Unlike Amazon Web Services or Microsoft Azure, which must balance databases, video streaming, and countless other workloads, Fluidstack markets itself as an AI-only “neocloud”—tailoring power, cooling, and interconnects exclusively for large-scale machine learning.
Imagem: Internet
The mechanics of an AI-only data center
Traditional enterprise data centers often mix CPU-heavy servers with limited GPU islands; Fluidstack instead deploys thousands of accelerators in a single hall, knitting them together with low-latency fiber to create enormous “training pods.” Cooling systems are optimized for the higher thermal output of GPUs, and electric supply contracts are structured to handle rapid load spikes when researchers launch multi-day training jobs.
This specialization can translate into meaningful performance and cost advantages for customers:
- Higher utilization because every rack is designed for machine-learning workloads.
- Faster interconnect thanks to purpose-built network fabrics.
- Energy efficiency through liquid cooling and renewable-energy sourcing agreements.
For AI companies racing to iterate on ever-larger models, shaving weeks off training cycles can be worth billions in potential revenue or first-mover advantage—explaining the premium valuations flowing into Fluidstack and peers.
Funding landscape: where the money might come from
Jane Street’s rumored lead role is noteworthy. The firm is best known on Wall Street for its algorithmic trading operations, yet its interest in infrastructure echoes a wider trend of hedge funds and specialist investors entering the AI supply chain. Other backers mentioned in prior rounds—Aschenbrenner’s Situational Awareness, the Collison brothers, Friedman, and Gross—reflect the convergence of AI researchers, entrepreneurs, and fintech founders rallying around compute scarcity.
Should Google decide to participate, the move would further blur lines between cloud vendors and capacity-hungry model developers. Alphabet’s cloud division already resells GPU and TPU compute to third parties; investing in Fluidstack could give it indirect access to additional capacity or potential future collaboration avenues.
What’s next?
Neither the size nor the valuation of the latest funding package is final, and Fluidstack declined to comment when contacted on Tuesday. Negotiations could still shift in scope or fall apart, particularly in a capital-intensive sector where construction delays and component shortages are common. Yet the mere prospect of a nine-figure raise at an $18 billion sticker price underlines just how aggressively investors are betting on physical infrastructure as the bottleneck to AI progress.
For now, all eyes remain on whether Fluidstack can convert preliminary term sheets into signed contracts—and then execute on the multibillion-dollar build-outs it has promised customers like Anthropic. If the company succeeds, the next generation of large language models and multimodal systems could be training inside facilities purpose-designed for their exponential appetites, miles away from the traditional cloud mega-parks that defined the last decade.
Either way, the race to secure GPUs, electricity, and real estate for AI workloads is accelerating, and Fluidstack is emerging as one of the most closely watched players in that high-stakes game.
Frequently Asked Questions
- What funding round is Fluidstack currently negotiating? The startup is in discussions to raise about $1 billion at a valuation close to $18 billion, according to Bloomberg.
- Why did Fluidstack’s valuation climb so quickly? A landmark $50 billion data-center deal with Anthropic and growing demand for AI-specific compute have fueled investor excitement, pushing the company’s valuation from $7.5 billion in December to a potential $18 billion this spring.
- Who are some notable backers and customers of Fluidstack? Reported or existing backers include Jane Street, Situational Awareness, the Collison brothers, Nat Friedman, and Daniel Gross. Customers range from Anthropic and Meta to Mistral, Poolside, and Black Forest Labs.
