Nscale IPO Stalls at $14B Valuation Gap as Investors Question Whether It Can Build for Anthropic

Nscale's chief executive is demanding a $30 billion valuation for the company's planned initial public offering while institutional investors are countering in the $16–19 billion range, Bloomberg reported — a gap of up to $14 billion that has stalled the offering and made the AI infrastructure sector's largest contracted compute deal into an argument about what that deal is actually worth.

The dispute is not primarily about Nscale's technology or its ability to build GPU clusters. It is about whether a $44.6 billion compute agreement with Anthropic constitutes a bankable financial asset when the data center campus required to fulfill it has not yet been funded — and investors are betting that it does not.

Read more: Anthropic signs $44.6B Vera Rubin compute deal with Nscale

Why a $44.6B Contract Doesn't Translate to $30B in Equity Value

The Anthropic agreement is structured in four tranches, each tied to a GPU cluster delivery milestone at Nscale's Monarch Compute Campus in Mason County, West Virginia, and specifying NVIDIA Vera Rubin NVL72 rack-scale GPU units. On its face, the total contract value of $44.6 billion makes Nscale appear to have one of the most valuable AI compute relationships in the industry.

The critical issue, as investors read the Nscale S-1 filing alongside the Anthropic agreement, is not a general cancellation right but the delivery-contingent termination provisions embedded in the contract's structure. Anthropic holds the right to terminate individual tranches if Nscale fails to deliver GPU clusters on schedule or fails to maintain specified uptime thresholds. Those termination rights are not a technicality — they are the mechanism through which the $44.6 billion headline figure could collapse to zero if Nscale cannot fund, build, and deliver on time. The contract runs generally non-cancelable through 2033, but its value is entirely contingent on delivery performance.

As of the S-1 filing, Nscale had not secured the construction financing needed to develop the Monarch campus. The $44.6 billion contract value is therefore contingent on capital Nscale does not yet have, a construction project that has not broken ground on its GPU infrastructure, and sustained uptime performance across a multi-gigawatt facility that does not yet exist. For investors applying standard project-finance analysis, that combination is a contingent asset — not a bankable revenue commitment. That gap between contracted value and buildable value is where the $14 billion dispute lives.

What Nscale's Own Filing Reveals

Nscale's S-1 provides context that complicates the CEO's $30 billion argument further. The company reported $140.6 million in revenue for the first half of 2026, annualizing to roughly $281 million — which at the $30 billion ask implies a valuation multiple of approximately 107 times annualized revenue. At the investor counter of $19 billion, the multiple falls to roughly 68 times, still an aggressive premium for a company posting a $1.02 billion net loss in the same six-month period alongside more than $8 billion in total debt.

The company also reports total contracted value of $103.4 billion across all customer agreements, but only $2.6 billion of that is currently active — meaning GPU clusters are delivered and generating revenue. The remaining $100.8 billion is contingent on future hardware delivery, power infrastructure completion, and construction financing. That 2.5% activation rate is not evidence of operational failure; it reflects the natural lag between signing compute agreements and building the facilities to honor them. But it means that Nscale's headline TCV figure overstates near-term revenue by a factor of roughly 40. Execution risk is already visible: Nscale's flagship facility in Glomfjord, Norway is running approximately six months behind schedule, with only 6.4 megawatts live against a 30-megawatt target, a shortfall that has already triggered customer penalties. Formal marketing of the IPO remains on hold pending the SEC's completion of its S-1 review.

CoreWeave Already Showed Investors How to Price a Neocloud at IPO

The most instructive precedent is CoreWeave's IPO in March 2025. CoreWeave — a structurally similar AI compute operator with major AI-lab and hyperscaler customers — targeted a $35 billion valuation and ultimately priced at approximately $19 billion on a shares-outstanding basis, a discount of roughly 46 percent from its initial ask.

Nscale's situation maps closely onto that template. Investors countering at $16–19 billion represent a discount of 37 to 47 percent from the $30 billion CEO ask — almost identical in magnitude to CoreWeave's pricing compression. What investors appear to be pricing is not Nscale's individual quality but a structural feature of the neocloud model: contracted TCV figures that cannot be treated as fixed-income equivalents, GPU delivery pipelines that create timing risk between signing and revenue realization, and debt structures that require IPO proceeds to remain serviceable. The precedent suggests capital markets have begun applying a systematic discount formula to neocloud operators with unbuilt infrastructure.

Monarch Campus: What Is Built and What Still Has to Be

The Monarch Compute Campus in Mason County, West Virginia is the physical foundation of Nscale's Anthropic relationship. The facility is designed around a behind-the-meter natural gas microgrid, giving Nscale direct control over energy supply and bypassing the utility interconnection queues that have delayed competing AI data center projects across the US. Nscale targets 1.35 gigawatts of capacity in the first phase across more than 2,250 acres in the Point Pleasant area, with an 8-gigawatt long-term buildout by 2031.

Currently, Nscale has 55 megawatts of live capacity across its global footprint. The Monarch campus, where the vast majority of contracted Anthropic capacity is to be deployed, has not yet broken ground on its GPU infrastructure. Microsoft's letter of intent for 1.35 gigawatts at Monarch, signed in March 2026, lapsed over the summer of 2026. Microsoft had earlier exited Nscale's Loughton, Essex site in the UK after power grid access was pushed into the 2030s. Those two Microsoft withdrawals, combined with the Norway delivery delays, mean that Nscale's most strategically consequential capacity remains on paper heading into its public market debut.

The Precedent That Extends Beyond Nscale

The Nscale stalemate is the first time a major neocloud operator with a publicly disclosed AI-lab mega-contract has undergone public price discovery at IPO while that contract's execution risk is simultaneously scrutinized by capital markets. Other AI compute platforms in earlier IPO planning stages are watching closely, because the outcome will determine how they structure future compute agreements with AI-lab customers and what level of operational execution investors will require before assigning full TCV credit.

If Nscale prices near the investor range, the market will have established that AI compute agreements need a more credible path to construction — secured financing, demonstrated delivery execution, and reduced distance between contracted and active TCV — before they can anchor an IPO valuation at a founder's preferred multiple. That is not merely a Nscale constraint. It is a signal to how Anthropic, OpenAI, and other AI labs will need to evaluate the financial health of their compute partners going forward. A compute partner that cannot raise capital to build is a supply chain risk.

The next milestone is not the IPO date — which remains undetermined — but whether Nscale can secure the financing to make the Monarch campus real. Until that happens, the $14 billion gap represents something more significant than a negotiation: it is the market's first published price for the execution risk embedded in the AI infrastructure build-out.