Nscale generated $140.6 million in revenue during the first half of 2026. It lost more than a billion dollars over the same period. On September 18, 2026, the London-based AI cloud and data center provider filed paperwork with the SEC to go public on the New York Stock Exchange anyway, targeting a valuation of up to $35 billion under the ticker symbol NSCL.

That gap between revenue and losses isn't a red flag investors are likely to treat as disqualifying. It's become the standard financial profile of an entire category of company racing to build physical AI infrastructure faster than any single balance sheet can comfortably fund.

Another Crypto Miner Turned AI Infrastructure Company

Nscale's origin story follows a pattern that's become familiar this year. The company spun off from a cryptocurrency mining business in 2024, repurposing infrastructure and expertise originally built for crypto into GPU-based AI computing capacity instead, the same pivot that took Crusoe from flaring waste gas at oil fields to building OpenAI's own Stargate campus. Both companies discovered that whatever infrastructure and power expertise they'd built for one computationally intensive industry translated surprisingly well into the other, once AI training and inference demand outpaced what traditional cloud providers could supply fast enough.

Since that 2024 spinoff, Nscale has scaled aggressively. The company now operates five active data centers containing 25,000 active GPUs, with 461,000 additional GPUs either active or already contracted, spread across hyperscale AI hubs in Norway, Portugal, Texas, and West Virginia, together representing more than 10 gigawatts of power capacity as of August 2026.

The Numbers That Explain Both the Hype and the Risk

Nscale's growth rate is genuinely extraordinary by any conventional measure. Revenue climbed 1,252% year over year, from $10.4 million to $140.6 million, during the first six months of 2026. Its net loss over the same period widened to $1.02 billion, up from $368.9 million a year earlier, meaning the company's losses grew even faster than its already explosive revenue. Nscale carries more than $8 billion in debt, and its prior private valuation, set in a March 2026 funding round, stood at $14.6 billion, less than half the valuation it's now targeting for this public offering.

The company points investors toward a different, forward-looking number to justify that valuation jump: $56.4 billion in remaining performance obligations, essentially contracted future revenue not yet recognized. That backlog figure is the crux of the entire investment case. Nscale is asking public market investors to value the company based substantially on contracts signed but not yet delivered, a bet that's become the defining characteristic of nearly every AI infrastructure company reaching public markets this year.

Nvidia Is Nscale's Supplier, Customer, and Backer All at Once

Nvidia's relationship with Nscale runs through nearly every layer of the business simultaneously. Nvidia has invested more than $2 billion directly in the company, supplies the GPUs that make up Nscale's core infrastructure, and has separately signed a $1.2 billion capacity leasing agreement to actually use Nscale's computing resources as a customer. Nvidia has also agreed to guarantee up to $860 million of Nscale's own debt obligations tied specifically to a Texas data center lease.

That's a remarkably concentrated set of roles for a single company to hold: investor, supplier, customer, and debt guarantor, all layered onto one relationship. It mirrors the kind of vendor-financing structure Nvidia has explored elsewhere in the AI infrastructure buildout, where the company backing a project's financing also stands to benefit directly from that same project's chip purchases, an arrangement that raises real questions about how independently any of these companies' growth numbers can actually be evaluated when so much of the underlying capital and demand ultimately traces back to the same source.

The Concentration Risk Sitting in the Prospectus

Nscale's own SEC filing discloses a customer concentration figure worth taking seriously: one unnamed customer accounted for more than half of the company's total revenue during the first half of 2026. That's an extreme reliance on a single relationship, and it means Nscale's actual growth trajectory depends heavily on whatever that one customer decides to do next, a genuine business risk hiding behind an otherwise blockbuster growth percentage.

To its credit, Nscale has taken visible steps toward diversification. The company counts OpenAI, Anthropic, and Microsoft, itself a competing cloud provider, among its named customers, and in July 2026 it announced plans to acquire Anyscale, a startup whose software helps other companies build and train their own AI models. That acquisition would push Nscale further up the AI infrastructure stack, from simply renting out GPU capacity toward offering the tools companies use to actually build models on top of that capacity.

A Board Built From Big Tech's Recent Departures

Nscale's board reads like a roster of recent high-profile departures from the largest AI and tech companies. Just last week, the company announced Fidji Simo, formerly OpenAI's product and business chief, would join its board. Simo stepped down from that OpenAI role in July after taking medical leave, a transition that reshuffled significant authority at OpenAI toward president Greg Brockman at the time. Nscale's board also includes former Meta executives Nick Clegg and Sheryl Sandberg, giving the company a genuinely deep bench of executives who spent years inside the exact large tech companies now driving the AI infrastructure demand Nscale is betting its entire business on serving.

What Nscale's Filing Says About the Neocloud Trade Right Now

Nscale joins a genuinely crowded field of AI infrastructure companies either already public or actively pursuing public listings this year, competing directly against established cloud giants like Amazon alongside fellow neoclouds including CoreWeave and Nebius. What distinguishes this specific filing is how transparently it lays out the core tension defining this entire category: massive, real revenue growth, massive, real losses, and a valuation that ultimately depends on investors trusting a backlog of contracted future business over the demonstrated financial performance sitting in the prospectus today.

That's the same underlying bet running through nearly every AI infrastructure story this year, from Nscale's own IPO filing to SK Hynix's $720 billion commitment to new memory fabrication capacity: that current losses and current debt are simply the cost of building capacity fast enough to meet AI demand that keeps arriving faster than anyone can supply it. Whether public market investors are willing to fund that bet at a $35 billion valuation for a company still losing more than a billion dollars every six months is a question this filing puts to the market directly, rather than one Nscale's own prospectus can answer on its own.