Nscale faces a $32 billion funding puzzle as losses balloon

Nscale has a revenue growth number most companies would frame and hang in the lobby. Its revenue is up 1,344% year over year.

The catch is the bill. The London-based AI infrastructure firm posted a $1.69 billion net loss and still lacks binding financing commitments for its biggest project, a GPU buildout tied to a contract with Anthropic.

That combination matters right now because Nscale is preparing to go public. Investors are about to be asked whether a company with a huge order book and a huge funding gap is a growth story or a construction loan with a ticker.

What the filing shows

The figures come from Nscale’s amended S-1 prospectus, which adds third-quarter results to the company’s earlier disclosures. Preliminary nine-month revenue reached $307.5 million.

That is strong growth by almost any standard. It is also a fraction of the $1.69 billion net loss over the same stretch.

Put plainly, Nscale lost more than five dollars for every dollar it brought in. For a company in heavy buildout mode, that is not unusual. For a company about to ask public markets for money, it is a number that will get circled in red ink.

The centerpiece of the Nscale story is its agreement with Anthropic, valued at $44.6 billion to $45 billion. Under the deal, Nscale would supply GPU capacity at its planned Monarch Compute Campus in West Virginia.

The initial spec for that campus is 460 megawatts of capacity, with potential room to expand. For context, that is the kind of power draw usually discussed in terms of small cities, not server rooms.

The $32 billion question

Building Monarch will not be cheap. Nscale estimates it needs $10 billion to $10.5 billion for data center and power infrastructure.

On top of that, it projects another $21 billion to $22 billion specifically for GPU equipment. Add the two ranges together and the tab lands somewhere between $31 billion and $32.5 billion.

Here is where the filing gets uncomfortable. Nscale cautioned that it has not secured binding commitments for that financing.

Think of it like a contractor who has signed a deal to build a skyscraper but has not yet lined up the loan to pay for steel. The contract is real. The building is not, at least not yet.

The company’s broader order book is large. Its total contracted value backlog stands at approximately $105.8 billion.

Backlog, though, is a promise of future revenue, not cash in the bank. Turning it into income requires the infrastructure to exist first, and the infrastructure requires capital Nscale says it has not yet locked down.

Progress on the ground

Nscale is not standing still while it hunts for financing. The company recently brought two additional data centers online.

Its active GPU count has grown to about 50,000 units. That sounds substantial until you compare it with the eventual target of 461,000.

In other words, Nscale is roughly a ninth of the way to where it says it wants to be. The remaining eight-ninths is where most of the capital spending, and most of the execution risk, lives.

Each step up that ladder means more chips, more power contracts, more construction, and more chances for something to slip. Supply chains, grid connections, and permitting do not typically move at the speed of AI demand.

The IPO pitch

Nscale is eyeing an initial public offering on the NYSE under the ticker NSCL. The company hopes for a market valuation between $30 billion and $35 billion.

That symmetry is not necessarily a problem, but it frames the pitch. Public investors would effectively be valuing the company on its ability to finance and deliver a project of roughly its own size.

Investor concerns about execution risk and the funding gap have raised questions about both the timing and feasibility of the listing. A company can have a blockbuster customer and still struggle to convince markets it can carry the weight.

What this means for investors and the AI buildout

For prospective IPO buyers, the core trade-off is visible in the filing itself. Revenue growth of 1,344% and a backlog of approximately $105.8 billion suggest real demand for what Nscale sells.

Against that sits a $1.69 billion net loss and no binding financing for a project with a combined capital requirement in the low-$30 billion range. The upside depends almost entirely on Nscale’s ability to raise money and then spend it well.

The Anthropic contract is both the strongest and the riskiest part of the story. A $44.6 billion to $45 billion commitment from a prominent AI developer is a powerful endorsement of demand. It also concentrates a large share of Nscale’s future on a single campus in West Virginia and a single customer relationship.

For the broader AI infrastructure market, Nscale is a useful stress test. Demand for GPU capacity has produced enormous order books across the sector, but those books only convert to revenue if someone funds the hardware and the power to run it.

If Nscale can close its financing and price its IPO near its target range, it would signal that capital markets are still willing to underwrite the AI buildout at scale. If it struggles, that would be a sign investors are starting to discount backlog that has not yet been paid for.

There are several things to watch from here. The first is any announcement of binding financing commitments for Monarch, which would directly address the biggest risk flagged in the filing.

The second is the GPU count. Movement from about 50,000 toward the 461,000 target would show whether Nscale can execute, not just sign contracts.

The third is the IPO itself. Whether NSCL prices within the $30 billion to $35 billion range, and whether it lists on the timeline the company hopes for, will say a lot about how much patience public markets have for companies that are growing fast and spending faster.