Nscale, the London-based cloud company built to rent computing power to artificial-intelligence labs, is talking to investors about raising as much as $3.5 billion before it goes public, according to people familiar with the matter. Bloomberg reported the discussions on Friday. The round would sit between the company’s existing venture funding and a stock-market debut that Nscale has said could come as soon as this year.
The sums involved have grown so large that they have stopped resembling startup finance. Nscale said this month that its contracted revenue, the value of compute deals already signed with customers, stands at roughly $103 billion. Much of that backlog is owed by the AI labs themselves: in late August the company agreed a contract with Anthropic worth up to $45 billion, one of a string of giant capacity deals the model builder has signed in recent months.
Nscale is a product of the same forces that created the AI infrastructure boom. The company, founded a few years ago by the British entrepreneur Joshua Payne, buys graphics processors by the tens of thousands, stands them up in purpose-built data centers and sells the resulting computing time to companies that cannot build their own. Its customers are largely the labs and startups whose models need more chips than any landlord can supply on short notice.
The company’s physical network has grown alongside its order book. Nscale operates data centers in Norway, including sites at Narvik and Glomfjord that run on renewable power, has a facility in Loughton outside London and has been developing sites in Texas. Its stated pipeline stretches past a gigawatt of greenfield capacity across Europe and North America, and it has announced plans to spend billions of dollars in Britain alone over the next several years.
The pre-IPO financing follows a familiar pattern in the AI infrastructure trade. Companies like Nscale, CoreWeave and their rivals sell future compute capacity under contracts that often run for years, then raise money against those contracts to pay for the data centers and chips the deliveries require. The scale of the commitments means the capital raising never really stops: each new contract creates the need for the next round of financing to fund it.
Investors in the round will be buying a claim on that backlog, and the question they face is how much of it will become cash. AI capacity agreements vary in enforceability. Some are take-or-pay arrangements that oblige the customer to pay whether or not it uses the compute; others carry cancellation clauses and escape hatches. Nscale’s own biggest contract, with Anthropic, was signed only weeks ago, and its durability will be tested by the very thing that created it: demand for AI compute that has repeatedly defied forecasters.
The company’s path to a listing has been the subject of speculation since the summer. Reports in August said Nscale was targeting a U.S. IPO as soon as September, a calendar that would make the current financing a final private raise rather than a bridge to somewhere far off. The company has not confirmed a date, and the size of the contracted revenue figure suggests the offering, when it comes, will be one of the largest in the young history of AI infrastructure listings.
Britain is watching closely. Nscale is the most prominent of the country’s AI cloud challengers, and its success or failure has become a proxy for whether Europe can host a serious AI compute industry rather than renting capacity from American providers. The government has courted data-center investment as a strategic priority, and Nscale’s expansion plans have been cited in that campaign.
The competitive field is crowded and getting more so. CoreWeave, the largest of the AI cloud specialists, is already public and expanding its own fleet. Hyperscale clouds from Amazon, Microsoft and Google are pouring capacity into the same customers. And the chip suppliers, above all Nvidia, have been lending and investing across the sector, collecting returns on both sides of the business: selling the processors and sharing in the revenue of the clouds that rent them out.
For Nscale, the $3.5 billion round is a test of investor appetite at a particular moment. AI infrastructure companies have commanded extraordinary valuations on the strength of contracted revenue that is large in nominal terms but spread over many years, and the market’s mood can shift quickly if doubts spread about the durability of AI demand. The company is asking private investors to price that risk before public markets get their turn.
People familiar with the company’s thinking say the money would go toward capacity, the same destination as every prior raise: more data centers, more chips, more power. Nscale’s customers have already committed to buy what it has not yet built, which is the fundamental wager of the AI cloud business. The company is now raising the capital to make good on those promises, and the round will tell the industry how much confidence the money behind AI has in the contracts that money is being spent to fulfill.
