LONDON — Josh Payne did not finish school. He spent years in the mining industry in Australia before pivoting to cryptocurrency infrastructure, then pivoted again to the only sector spending faster than crypto had in its peak years. The company he built from those pivots filed for a listing on the New York Stock Exchange three days ago, targeting a valuation of $50 billion.
Nscale Global Holdings Ltd. submitted its Form S-1 registration statement with the Securities and Exchange Commission on September 18, seeking to trade under the ticker NSCL. The company is targeting a raise of up to $3 billion. In March, Nscale was worth $14.6 billion. The gap between those two numbers in six months is not a rounding error.
The raw financials are striking. Revenue in the first half of 2026 hit $140.6 million, up 1,252 percent from $10.4 million in the same period a year earlier. The net loss in the same window was $1.02 billion, more than double the $368.9 million Nscale lost in the first half of 2025. Nvidia, a Series B backer alongside Dell and Nokia, separately forecasts that AI chip demand will double next year. The company burning through capital and the chip supplier funding it are operating on the same underlying thesis: this market is not yet close to the scale it will reach.
The losses have a cause. Nscale is not a software business. It is the physical layer. It builds and operates data centers, deploys GPU fleets, and sells compute capacity to the AI companies that need more of it than they can currently commission elsewhere. Its power pipeline exceeds ten gigawatts. As of August, it had 25,000 GPUs active and another 461,000 active or under contract. Five data centers are live, and twelve more are contracted. The money being spent right now is buying capacity that customers have already committed to rent, often for years.
What that reporting also noted, and what Nscale’s S-1 does not foreground: Microsoft had previously held a letter of intent for that same West Virginia location, with plans for up to 1.35 gigawatts. Google was in discussions too. Both walked away before the summer was over. The facility that the two largest cloud platforms declined went to one of their main AI competitors for forty-five billion dollars.
That sequence is either a cautionary detail for Nscale’s prospective shareholders, since major customers are mobile, or evidence that Nscale occupies a position neutrally available to whoever bids highest. The company Payne co-founded with Nathan Townsend in May 2024 started as an offshoot of Arkon Energy, a crypto mining infrastructure vehicle. When GPU demand pivoted from proof-of-work mining toward AI training, the physical assets transferred cleanly. The Series A was $155 million. The Series B was $1.1 billion, with Nvidia coming in alongside Dell, Nokia, and Aker. The Series C, in March 2026, raised $2 billion at a $14.6 billion valuation. The time between Series C and S-1 filing is six months.
Goldman Sachs, J.P. Morgan, and Morgan Stanley are leading Nscale’s offering. The company’s S-1 does not specify pricing or a timetable, but it reports more than $88 billion in committed multiyear revenue from anchor customers including Anthropic and Microsoft.
Those figures represent existing contractual commitments, not projections.
Nscale and Anthropic are also moving through the same IPO window. Anthropic, which accounts for the $45 billion anchor contract, is reportedly targeting a New York Stock Exchange debut in November at a valuation of approximately $2 trillion.
The offerings represent opposite sides of the AI infrastructure equation: Nscale provides the computing capacity, while Anthropic consumes it.
Meta, Apple, and Google are all building their own chip infrastructure now. The companies currently renting capacity from Nscale may, in time, need less of what Nscale is selling. Payne’s company has not answered that question publicly. The S-1 does not try to. What it offers instead is the trajectory of the last eighteen months, and asks investors to extrapolate.
