Fluidstack seeks $1 billion investment valuing AI startup at $18bn with Jane Street backing

Fluidstack, the specialised AI infrastructure startup, is in advanced discussions to raise approximately $1 billion in a funding round that would value the company at around $18 billion, according to people familiar with the matter. If finalised, the deal would represent a dramatic valuation surge for the company, which was reportedly seeking a $7.5 billion valuation as recently as December last year.

Morgan Stanley is advising on the deal, the terms of which remain subject to change. The round could be co-led by the trading firm Jane Street and the AI-focused hedge fund Situational Awareness, which manages over $1.5 billion and was founded by former OpenAI researcher Leopold Aschenbrenner.

The $50 Billion Foundation of a Valuation

The staggering leap in valuation is largely anchored by a single, landmark agreement. In November 2025, Fluidstack and the AI lab Anthropic announced a $50 billion deal to develop custom AI data centres in the United States, with initial sites in Texas and New York and more planned. This partnership, custom-built for Anthropic’s specific workloads and research needs, underscored a seismic industry shift: leading AI companies are moving beyond renting standard cloud capacity to ordering dedicated, optimised infrastructure, seeking greater control over costs, performance, and scaling.

This demand is the core of Fluidstack’s business. The company operates as a “neocloud” provider, a newer category focused on supplying large clusters of GPUs and building dedicated computing capacity, often by converting former Bitcoin mining sites or other power-heavy industrial locations.

Accelerating AI by Rethinking the Cloud

Fluidstack’s fundamental proposition is that it accelerates AI projects by removing compute bottlenecks in a way traditional hyperscalers like AWS, Google Cloud, or Microsoft Azure do not. Where those providers cater to a vast array of general computing needs, Fluidstack focuses exclusively on infrastructure fine-tuned for the intense demands of AI training and inference workloads.

The company employs a dual-pronged B2B model. One revenue stream is a marketplace platform connecting customers to third-party GPU capacity. The other, higher-margin segment is a private cloud where Fluidstack owns and operates its own GPU infrastructure, offering AI firms deeper control. The company manages a network of over 100,000 GPUs, with its model creating significant new revenue streams for partners like former Bitcoin miners.

This specialised approach contrasts with the broader service suites of traditional clouds and addresses a critical pain point as AI models grow exponentially in size and complexity. The model has proven attractive to AI labs that require predictable, high-performance compute at scale, free from the potential constraints of multi-tenant cloud environments.

Fluidstack’s strategic focus has narrowed sharply onto the American market. Originally founded in Europe, the company relocated its headquarters to New York City in December 2025. It has stepped away from international projects to prioritise U.S. expansion, including withdrawing from a planned €10 billion AI infrastructure project in France to concentrate on opportunities tied to American clients like Anthropic.

A Client Roster of AI Heavyweights

The company’s client base reads as a who’s who of the generative AI sector, underlining its position as critical infrastructure for the industry. Beyond its anchor tenant Anthropic, Fluidstack’s customers include Meta, the software development AI firm Poolside, the French AI leader Mistral, Character.AI, the image generator Midjourney, and Black Forest Labs.

This client concentration highlights both the opportunity and the risk in Fluidstack’s model. Its business is heavily dependent on the continued breakneck growth and demand of a small number of elite AI companies, and it is inextricably linked to the supply and dominance of Nvidia’s GPU hardware. The rapid valuation increase it now seeks mirrors the immense premiums investors are placing on specialised AI infrastructure, as capacity constraints at larger cloud providers persist and the race for computational power defines the AI era.