British AI infrastructure company Nscale is preparing to test public-market appetite for the expensive cloud infrastructure behind generative AI after filing for a U.S. initial public offering.
The London-based company filed its Form S-1 with the U.S. Securities and Exchange Commission on September 18 and plans to list on the New York Stock Exchange under the ticker NSCL. The number of shares and expected IPO price have not yet been determined.
Reuters reported that Nscale’s first-half revenue increased 1,252% to $140.6 million, compared with $10.4 million a year earlier. But its net loss also widened sharply, reaching $1.02 billion, from $368.9 million during the same period in 2025.
Nscale is building around power as much as compute
The filing provides a useful look at what competing in AI cloud infrastructure now requires.
Nscale says it has expanded from 750 megawatts to more than 10 gigawatts of owned and controlled power capacity, while growing from 40 employees to more than 1,000 and operating across 14 regions. Its total contracted value has expanded from $100 million to more than $103 billion in roughly two and a half years.
That scale reflects a broader shift in the AI infrastructure market. Access to GPUs remains critical, but Nscale argues that large-scale contiguous power has become a primary constraint on AI infrastructure deployment.
Its strategy therefore reaches beyond renting computing capacity. Nscale develops data centers, power infrastructure and GPU clusters alongside the cloud software used to operate AI workloads.
Big contracts come with concentration risk
The company has already secured unusually large commitments.
As of August 31, Nscale reported approximately $103.4 billion in active and contracted total contract value, supporting around 461,000 active or contracted GPUs. Its contracts had a weighted average life of roughly 5.7 years.
Reuters reported that Anthropic agreed last month to spend about $45 billion on computing capacity at Nscale’s West Virginia campus. Microsoft and Anthropic are expected to become major customers in future periods.
The concentration also creates risk. Nscale’s largest customer represented 52% of revenue in the first half of 2026.
Infrastructure growth is consuming enormous capital
Nscale’s expansion illustrates why the AI cloud market increasingly resembles infrastructure finance rather than conventional SaaS.
The company disclosed major debt facilities for GPU and data-center projects, including $1.2 billion for North Carolina and $1.85 billion for a Texas facility. It also says it expects significant infrastructure investment to continue.
The IPO therefore tests something larger than investor appetite for another AI company.
Nscale is competing with CoreWeave, Nebius, Crusoe and Lambda in a market where securing models is not enough. Providers increasingly need GPUs, software, cooling, financing, data centers and gigawatts of power simultaneously.
For European technology, that marks a significant evolution. The continent’s AI competition is moving beyond model development into the physical and financial infrastructure required to keep those models running at industrial scale.
