Nscale Files for NYSE IPO Targeting $30 Billion Valuation, Reports $1.02 Billion Loss on $140.6 Million Revenue

Since Nscale filed confidentially with the SEC on February 17, 2026, and was valued at $14.6 billion in a March funding round, the London-based AI data center builder has more than doubled its target valuation to roughly $30 billion in the public S-1 registration statement it submitted September 18 for a New York Stock Exchange listing under ticker NSCL.

The filing sets up one of the largest AI infrastructure listings since CoreWeave went public in March 2025. Goldman Sachs, J.P. Morgan and Morgan Stanley lead a syndicate of 23 underwriting banks, according to Value Add VC's review of the filing. No share count or price range has been disclosed yet. Renaissance Capital estimates the raise at roughly $2.0 billion; Tech Times reports the company is targeting $3 billion, double what CoreWeave raised in its own IPO.

The Revenue Gap

Nscale reported $140.6 million in revenue for the first six months of 2026, a 1,252% jump from $10.4 million a year earlier, according to Channel NewsAsia's review of the filing. But the net loss grew even faster, hitting $1.02 billion in the same period, up from $368.9 million a year prior.

The company's headline figure is $103.4 billion in total contract value, or TCV, as of August 31, 2026. That number reflects long-term take-or-pay commitments from customers, not cash collected. Value Add VC calls the gap between that backlog and actual recognized revenue "the risk sitting under this entire filing." Tech Times frames it more bluntly, calling the ratio between $103.4 billion contracted and $140.6 million recognized "the defining characteristic of the neocloud business model" that any investor reading only the headline number is pricing wrong.

Under that model, Nscale takes a signed multi-year contract from a customer like Anthropic or Microsoft, borrows against the guaranteed future cash flow, buys Nvidia chips, builds the data center, and only starts booking revenue once capacity is delivered and running.

Anthropic Is Nearly Half the Backlog

A single deal with Anthropic, announced in late August, accounts for 44% of Nscale's entire $103.4 billion backlog, according to Value Add VC. That agreement runs five years and covers 460 megawatts at Nscale's planned West Virginia campus, valued at $45 billion.

Microsoft and Anthropic together represent up to roughly $88 billion in multi-year commitments, per Renaissance Capital's filing review. Channel NewsAsia reports that Nscale's single largest customer accounted for 52% of its first-half revenue. Anthropic told investors, according to Fox News' reporting citing New York Times Dealbook, that it wants roughly five gigawatts of computing power online by the end of 2026 and plans to double that to 10 gigawatts by the end of 2027. This buildout pace helps explain why Anthropic has signed roughly five separate multi-billion-dollar compute deals this year with AWS, FluidStack, Nscale, SpaceX and Lambda.

That concentration cuts both ways. It shows real demand from a fast-growing AI lab. It also means Nscale's fortunes are tied heavily to whether Anthropic's own spending plans and eventual IPO ambitions hold up.

Nvidia Is Both Supplier and Financier

This week, Nscale sold $3.1 billion in convertible bonds, including $1 billion purchased directly by Nvidia, according to Channel NewsAsia. Value Add VC notes this is now a recurring pattern across the neocloud category. Nvidia invests directly in companies that then use that capital to buy Nvidia chips. The open question is whether that's Nvidia underwriting genuine demand for its hardware or simply backing the fastest distributor of it.

Nvidia's ubiquity in this space stands in some tension with a separate development. Breitbart reported that TIME's 2026 list of the 100 most influential people in AI dropped Nvidia CEO Jensen Huang entirely, after he had appeared on every prior edition and shared TIME's 2025 Person of the Year honor with the company. Nscale co-founder Josh Payne made this year's list; Huang did not.

The Bull and Bear Case

Matt Kennedy, senior strategist at Renaissance Capital, told Channel NewsAsia that "the setup for AI infrastructure is good enough to get these deals done, but it's nothing like the euphoria of a few months ago." He noted neocloud players like Nscale, CoreWeave and Nebius "seem to be holding up better than the AI infrastructure stocks that sit further up the supply chain," pointing to Nebius's recent price hikes and CoreWeave's ability to sign contracts at higher prices as evidence of real demand.

That framing matters because, according to Channel NewsAsia, the listing comes weeks after top AI lab CEOs called for a slowdown in AI development, a comment that rattled markets. Kennedy's read is that despite that volatility, "this space is still attracting billions" because of genuine bottlenecks in AI compute and power.

The skeptical case, laid out by both Tech Times and Value Add VC, is that a $103.4 billion backlog built substantially on one $45 billion contract with a company that hasn't gone public itself is a customer-concentration bet dressed up as an infrastructure bet. If Anthropic's compute demand slows or Nscale's buildout timeline slips, the backlog doesn't automatically become cash flow.

Sources also disagree on Nscale's current GPU footprint. Tech Times reports 25,000 GPUs actively generating revenue with 461,000 more active and contracted for future delivery, while Value Add VC's review of the filing puts the active-and-contracted total closer to 289,000, including about 194,000 Nvidia Vera Rubin chips. Neither figure has been independently reconciled against the other in the filing summaries reviewed.

No pricing terms have been set, and Nscale has not disclosed a roadshow date. The company competes directly against Crusoe, which closed a $3.9 billion Series F at a $30.9 billion private valuation the same week Nscale filed, according to Value Add VC. That suggests Wall Street's appetite for AI infrastructure bets, public and private, hasn't cooled nearly as much as some AI executives' recent slowdown warnings suggested.