NextFin News - Nscale, the UK-based AI infrastructure builder backed by Nvidia and Microsoft, is working with Goldman Sachs and JPMorgan on a US initial public offering that could raise as much as $3 billion, according to people familiar with the matter. The potential listing, which could come as soon as September 2026, would make Nscale the second major "neocloud" provider to tap public markets after CoreWeave's landmark 2025 debut — and the clearest test yet of whether Wall Street will pay a premium for companies that own the physical backbone of the AI boom rather than the models running on it.
The Deal and the Pitch
Nscale is telling prospective investors it has approximately $51 billion in contracted revenue, with second-quarter 2026 revenue exceeding $100 million. Those figures, if verified in a registration statement, would position the company as one of the most heavily contracted private technology companies to come to market in years. The company is privately valued at $14.6 billion after a $2 billion Series C round in March 2026, meaning a $3 billion IPO would represent a significant step-up in scale even before any public-market premium.
The timing is deliberate. CoreWeave, the closest comparable, went public in March 2025 and raised $1.5 billion at a $23 billion valuation; its shares closed at $93.17 on August 18, 2026, more than double the $40 IPO price, after touching an all-time high of $183.58 in June 2025. That performance has effectively opened an IPO window for the entire neocloud category — specialized cloud operators that own and operate GPU-heavy data centers built specifically for AI workloads, rather than retrofitting generic facilities.
Nscale's pitch rests on contracts that look hyperscaler-grade. In October 2025 it signed an expanded agreement with Microsoft for approximately 200,000 Nvidia GB300 GPUs across Europe and the United States — one of the largest AI infrastructure contracts ever signed, according to the company. The deal spans a roughly 240-megawatt hyperscale AI campus in Texas, leased from Ionic Digital, where phased delivery begins in the third quarter of 2026, plus a site in Sines, Portugal. Nscale plans to scale the Texas footprint to 1.2 gigawatts over time, with Microsoft holding an option on a second 700-megawatt phase starting in late 2027.
The company's rise has been unusually fast even by the standards of the AI buildout. Spun out of crypto-mining operator Arkon Energy in 2024, Nscale raised a $1.1 billion Series B in September 2025 — the largest in European history at the time — followed by a $433 million pre-Series C SAFE round and then the $2 billion Series C in March 2026 led by Norwegian energy group Aker ASA and 8090 Industries. The roster of backers reads like a roll call of the AI supply chain: Nvidia, Dell, Nokia, Citadel, Jane Street, Point72, and Fidelity, with Goldman Sachs and JPMorgan acting as joint placement agents on the Series C — the same two banks now steering the IPO.
Why the Neocloud Model Exists — and Why It Commands a Premium
The neocloud thesis is simple but powerful: hyperscalers cannot build AI-specific capacity fast enough, and AI labs do not want to wait. Traditional cloud providers optimize for general-purpose workloads; neoclouds optimize for one thing — dense clusters of Nvidia GPUs networked for large-scale training and inference. That specialization matters because the economics of AI are dominated by GPU utilization, power cost, and time-to-deploy, not by the breadth of services on offer.
Nscale's contract book illustrates the model's appeal. The $51 billion in contracted revenue the company is reportedly sharing with investors is a backlog metric — revenue promised under long-term agreements — and it is roughly 3.5 times the company's $14.6 billion private valuation. For comparison, CoreWeave went public with a $25.9 billion backlog against a $23 billion valuation, a ratio of about 1.1 times. The gap is the heart of Nscale's marketing story: investors are being asked to underwrite capacity expansion at a backlog multiple that CoreWeave only reached after its shares had already doubled.
The energy angle is the second pillar. Nscale's facilities are anchored in Europe, where it says it can access some of the lowest-cost renewable power in the world — Norway, Iceland, Portugal — and it has pledged waste-heat reuse and local skills development as part of its social license to build. Power is the binding constraint on AI data centers, not chips; Nvidia's GB300 racks draw far more electricity per unit than previous generations, and grid interconnection queues in the US routinely stretch to years. A provider that can secure cheap, reliable, carbon-light power and connect it to the grid faster than incumbents has a structural cost advantage, not a cyclical one.
The sector's growth backdrop supports the timing. Neocloud revenues exceeded $25 billion in 2025 and reached $9 billion in the fourth quarter alone, up 223% year over year, according to Synergy Research Group, which forecasts the market will approach $400 billion by 2031 at a 58% compound annual growth rate. Nscale is not arriving in an empty market — it is arriving in one that investors have already decided is real.
Cyclical or Structural? A Structural Asset Wrapped in a Cyclical Valuation
This is the central question for any investor in the AI infrastructure buildout, and the answer is not uniform. The neocloud sector's growth is structural: AI compute demand is being driven by a permanent shift in how software is built and run, and the physical constraints of power and grid capacity will not self-correct on a quarterly cycle. A data center permitted and powered today will remain an income-generating asset for a decade or more, regardless of where the AI hype cycle trades.
But the valuation attached to that infrastructure carries a cyclical overlay. GPU prices, utilization rates, and the willingness of hyperscalers and labs to sign multi-year take-or-pay contracts all fluctuate with the AI capital-expenditure cycle. If AI spending slows — or if Nvidia's next-generation chips make existing GB300 clusters economically obsolete faster than expected — contracted revenue can prove less durable than the headline number suggests. The distinction matters: the assets are long-lived, but the pricing power embedded in today's contracts is not guaranteed to persist.
The honest read is that Nscale is a structural asset wrapped in a cyclical valuation. The facilities, power agreements, and grid connections are the durable part. The multiple investors are willing to pay for them is the part that mean-reverts.
The Second-Order Question the Market Hasn't Priced
The obvious story is that Nscale's IPO validates the neocloud model. The second-order story is more uncomfortable: a successful $3 billion raise would accelerate a supply response that erodes the very scarcity premium neoclouds enjoy today.
Every neocloud IPO signals to the market that AI infrastructure is a fundable asset class. That attracts capital — debt, equity, and project finance — which funds more data centers, which increases total GPU supply, which over time pushes down the price of AI compute. The companies best positioned to benefit are not necessarily the neoclouds themselves, but their customers: the AI labs and hyperscalers whose cost per token falls as capacity expands. In other words, the IPO that enriches Nscale's early investors also subsidizes the competitive intensity that will pressure Nscale's margins in 2028 and 2029.
There is also a customer-concentration risk that the market has barely discussed. CoreWeave derived 72% of its revenue in the first half of 2025 from a single customer, widely understood to be Microsoft, before diversifying into OpenAI and Meta contracts. Nscale's reported $51 billion in contracted revenue is heavily anchored to Microsoft and, to a lesser extent, OpenAI through the Stargate Norway joint venture with Aker. If one anchor customer renegotiates, delays deployment, or shifts strategy, the backlog — the company's single most important metric — can unwind faster than it was built. CoreWeave's playbook of layering multiple hyperscaler contracts is the template Nscale will need to follow, and it has not yet demonstrated it can do so at scale.
The Counter-Thesis — Why This Could Be a Top
The strongest argument against Nscale's IPO is not that the company is weak; it is that the window may be closing. Bears would point to three facts. First, CoreWeave's stock, while more than double its IPO price, has fallen roughly 50% from its all-time high near $184 in June 2025 — a sign that the initial euphoria around AI infrastructure equities has already cooled. Second, the IPO is being marketed on contracted revenue, a metric that is only as good as the counterparty's willingness and ability to perform over a multi-year horizon. Third, Nscale's only public financials — covering the early stages of the business through December 2024 — showed a $24 million loss, and the company is piling on debt to fund its buildout, including a $1.4 billion GPU-backed loan in February 2026 and a $790 million Norwegian facility in May 2026.
The bear case is coherent: Nscale is selling equity at the peak of the AI capital-expenditure cycle, using a backlog metric that may not survive a downturn, while loading the balance sheet with debt that must be serviced regardless of utilization. If AI spending growth decelerates in 2027, Nscale could face the worst of both worlds — falling revenue per GPU and rising interest costs on a heavily leveraged balance sheet.
The rebuttal is that the contracts are the point. Long-term, take-or-pay agreements with investment-grade counterparties convert a cyclical demand stream into annuity-like cash flow. If Nscale can prove in its registration statement that its contracts contain minimum-commitment clauses and that its power costs sit below the marginal producer's, the company is not a bet on AI hype — it is a bet on the spread between contracted revenue and the cost of capital, a spread that remains positive even in a slower-growth environment.
What Comes Next
The near-term path is clear: Nscale will file a registration statement, likely in the coming weeks, and begin marketing the deal to institutional investors with Goldman Sachs and JPMorgan leading the book. The success or failure of that marketing effort will hinge on two numbers: the verified size of the contracted-revenue backlog and the valuation at which the company is willing to sell. A pricing at or modestly above the $14.6 billion Series C valuation would signal a receptive market; a deep markdown would suggest the window is narrower than the company hopes.
Short term, the IPO is a sentiment event for the entire AI infrastructure complex. A strong debut would lift private valuations across the neocloud sector and encourage other candidates — Lambda is reported to be working with Morgan Stanley, JPMorgan, and Citigroup on a listing, and Crusoe has held IPO talks with four Wall Street banks — to file. A weak reception would have the opposite effect, freezing the private market and forcing down-rounds.
Medium term, the fundamental question is execution. Nscale must deliver the Texas and Portugal campuses on schedule, convert contracted GPUs into billable capacity, and diversify beyond its Microsoft anchor. The company's management team has built more than 50 data centers, and its board now includes Sheryl Sandberg, Susan Decker, and Nick Clegg — a governance upgrade that speaks to the scrutiny a public company will face.
Long term, the structural case stands: AI compute demand is not reverting, and purpose-built infrastructure with secured power has durable value. But the valuation multiple is cyclical, and it will compress as supply catches up. Investors should treat Nscale not as a pure AI growth story, but as an infrastructure yield story wrapped in a growth multiple — and watch the spread between the two.
"We have public market ambitions, and execution is an enormous focus of mine," Josh Payne, Nscale's chief executive, said in October 2025, when he first signaled a listing in the "back end of next year."
The falsifying signal: if Nscale's registration statement shows contracted revenue materially below the reported $51 billion, or if CoreWeave's stock falls below its $40 IPO price before Nscale prices its deal, the thesis that the neocloud window is open is wrong.
Nscale's IPO is less a bet on artificial intelligence than a bet on who owns the meter — and the meter owners only win if the power stays cheap and the contracts hold.
Explore more exclusive insights at nextfin.ai.
