Models & Labs · United States

Nscale's IPO filing pairs big AI contracts with warnings on delivery

Credited to Rackline · Sarah Chen · sec.gov · 1h ago

Photograph · Rackline · Sarah Chen

Nscale's prospectus lists Microsoft work worth up to about $43.8bn and Anthropic agreements worth up to about $44.6bn and, a few pages later, explains how they could unravel. The London-based AI cloud company filed to list on the New York Stock Exchange on Sept. 18. Its risk factors say that if it misses agreed delivery dates at its West Virginia campus, Anthropic "may be entitled to terminate that tranche without liability."

Nscale filed a registration statement on Form S-1 for an initial public offering of ordinary shares under the ticker NSCL, the company said in a Sept. 18 release. It has not set the number of shares or a price range. Goldman Sachs, J.P. Morgan and Morgan Stanley are lead bookrunners. The filing followed four confidential drafts submitted from February through August, according to the company's filing index at the Securities and Exchange Commission.

The contracts come from two buyers. Between September 2025 and April 2026, Nscale signed statements of work with Microsoft that provide for payments of "up to approximately $43.8 billion through December 2033," the prospectus says (p. 27). On Aug. 25 it signed four agreements with Anthropic for "aggregate payments to us of up to approximately $44.6 billion." Both sums depend on Nscale meeting delivery and availability requirements.

The Anthropic work is to be delivered at the Monarch Compute Campus in Mason County, West Virginia, which Nscale gained in March when it bought American Intelligence & Power Corp. The prospectus describes a site of about 2,250 acres that "can deliver up to 8GW of gross behind-the-meter power," and says Nscale expects "to develop up to 2 GW of gross power generation capacity by the first half of 2028" (p. 9).

Those dates carry penalties. "The Anthropic Services Agreements provide only limited relief in the event of supply chain delays, and if we fail to meet agreed delivery timelines for any tranche, Anthropic may be entitled to terminate that tranche without liability," the filing says (p. 26). It adds that "certain of our other customer contracts provide the applicable customer with a right to terminate for delays."

The agreements also include "delivery milestones, as well as stringent uptime and availability requirements, together with termination rights and other contractual remedies in the event of delay or sustained underperformance," according to the prospectus (p. 28).

Funding for Monarch is not yet committed. The agreements require Nscale to use best efforts to obtain qualifying financing for the GPUs and infrastructure there within a specified period. "As of the date of this prospectus, we have not obtained binding commitments for any of the financings required to fund performance under the Anthropic Services Agreements, and all such financings remain in process," the company says (p. 27).

Construction is a risk in its own right. "The expansion of our data center infrastructure, including new data centers, will be complex, and delays in the completion of these projects or the cost, availability or access to components necessary for these projects may result in increased expenditures, operational and project inefficiencies, delays, or interruptions in the delivery of our services to our customers," the filing warns (p. 23). It does not report a delay at any named site.

The operating business is small next to the contracts. Nscale reported revenue of $140.6 million for the six months to June 30, up from $10.4 million a year earlier, and a net loss of $1.02 billion, against $368.9 million (p. 18). For all of 2025, revenue was $33.0 million and the net loss $761.8 million.

Revenue is concentrated. The largest customer accounted for 52% of revenue in the first half of 2026 and 73% in 2025, the filing says (p. 27); in 2024 a single customer accounted for substantially all of it. Nscale says it expects Microsoft and Anthropic to be significant customers in future periods.

The footprint is mostly still to be built. As of Aug. 31, Nscale counted about 25,000 active GPUs and 461,000 active and contracted GPUs, five active and twelve contracted data center sites, and about 1.37 gigawatts of active and contracted capacity, 1 gigawatt of it at sites it owns, the prospectus says.

Portugal is one of the newer pieces. On April 24, Nscale bought SIN02, a 200 MW powered land site at the Sines data campus. The prospectus says the project will support its work with Microsoft "through the deployment of over 66,000 NVIDIA Vera Rubin NVL72 GPUs starting in late 2027" (p. 11).

The European plan has already changed once. Nscale announced Stargate Norway in July 2025 and Stargate UK in September 2025, with OpenAI among the partners. "In April 2026, OpenAI withdrew from the Stargate Norway and Stargate UK partnerships, with Microsoft subsequently expanding its agreement with us to take the extra compute capacity at Stargate Norway," the filing says (p. 37).

The offering itself is not earmarked for Monarch. Nscale says it intends to use the net proceeds "for general corporate purposes, including to fund and support our data center projects and deployments, technology development, working capital and operating expenses," and warns that "our use of these proceeds may differ substantially from our current plans" (p. 66).

Until then the company is borrowing site by site. The prospectus lists a $725 million term loan and a $65 million VAT facility for the Kvandal South data center in Norway, signed in July (p. 60); a $331.9 million facility from Macquarie for Iceland; up to $1.85 billion under a GPU facility for Ward County, Texas; and up to $1.2 billion under a North Carolina GPU facility. It also reports about $2.54 billion of initial-term rent under equipment framework agreements as of Sept. 4.

Equity has come in tranches. On March 27, Nscale issued 99.6 million Series C preferred shares for $1.64 billion, with new shares priced at $20.39 each (p. 95). On Sept. 15 it signed agreements for at least $3.1 billion more: $2.1 billion of unsecured convertible loan notes and a further $1.0 billion of notes or non-voting shares to be issued to Nvidia (p. 13).

Nvidia is supplier, customer, investor and guarantor. Nscale provides reserved GPU capacity to Nvidia under a master cloud services agreement worth $1.2 billion over initial terms of 12 to 48 months, and Nvidia agreed to guarantee up to $860.3 million of Nscale's obligations under its Ward County lease in exchange for warrants, the filing says (p. 145). Three suppliers accounted for 53%, 37% and 10% of related purchases in the first half (p. 24).

Josh Payne, the founder, chair and chief executive, sets out the pitch in a letter in the filing: "From inception, we built Nscale with an infrastructure-first thesis, building against contracted customer demand, underwriting projects to attractive long-term returns, maintaining prudent leverage and seeking to match the duration of our capital commitments with the strong revenues supporting them."

The structure will change before listing. A reorganization in May made Nscale Limited, registered in England and Wales, the holding company of Arkon Energy and Nscale Global Holdings. Before the offering it will re-register as a public limited company named Nscale plc.

The order of events is what investors will weigh. Most of that money is payable only if Nscale delivers capacity on time; delivering at Monarch requires financing the company says it has not secured; and missed dates can let the customer walk away from a tranche. The next marker is an amended filing with a share count and price range, which would show how much of the build the offering is meant to fund.

Models & LabsUnited States

OpenAI Poaches Another SpaceX Sales Lead Focused on AI

Credited to Bloomberg$ · bloomberg.com1h ago

Useful

OpenAI has hired SpaceX’s Graham Seamans to lead sales for the firm’s cybersecurity business, the latest in a string of hires from Elon Musk’s artificial intelligence venture.

View the article · $Bloomberg

Front Method

Models & Labs · United States

Replit is no longer working with Anthropic, CEO says

Credited to TechCrunch · techcrunch.com · 1h ago

Replit, the AI coding platform that has become a favorite of startup founders and developers, is no longer working with Anthropic, the company’s CEO and co-founder Amjad Masad said Thursday while onstage at TechCrunch Disrupt.

Masad said the decision was made for a number of different reasons, including cost. Replit announced a partnership with OpenAI in August, but there was never any indication that its was no longer with Anthropic.

“I think ChatGPT is better now,” Masad said, adding that company uses an internal benchmark to help assess metrics like performance. “We publish a public benchmark as well, and consistently, Sol and Astra perform better on building vibe-coded applications from scratch. It was it was a decision based on both cost, you know, partnership, and performance. And for all these reasons, we just felt like they were a better partner for now.”

Replit sells a cloud-based software development platform where users can develop and deploy apps directly from inside their browser. Under its partnership with Anthropic, Replit used the AI lab’s Claude models to run its platform. That partnership was said to have helped the company grow its revenue significantly.

Masad’s comments were met with some pushback from insiders familiar with the partnership between the Anthropic and Replit, however. One person familiar, who asked not to be named, said the partnership was still in place.

There is some evidence that Replit was using Anthropic products as recently as September. For instance, a blog post in September cites Anthropic’s Claude Fable 5 and Claude Fable 5.1 as models running in Replit Agent’s production environment.

TechCrunch reached back out to Replit for clarification and will update this story if we hear back.

Masad shared insights about the company’s growth while onstage at Disrupt, noting that enterprise is starting to provide a larger slice of revenue.

“We’re not sharing ARR numbers just yet, but in terms of growth, like, I would say, enterprise was like 2% of revenue last year,” Masad said. “This year, I would say like 15% of our revenue is enterprise contracts, and more than 50% of our revenue is B2B.”

Masad also commented on the necessity for good AI safety measures.

“In the same way that us humans need a certain environment to work — you need a nice office, maybe nice good food in the office, air conditioning, computer — agents need a habitat for them to exist and do good work,” he said. “So that habitat has good orchestration, good error recovery, good safety and security and sandboxing, good networking, good identity and authorization. And once you figure all of that out, then you can deploy AI safely.”