Nscale, the London-based AI cloud company backed by Nvidia, has filed a registration statement with the SEC for an initial public offering on the New York Stock Exchange under the ticker NSCL. The filing, dated September 18 and picked up widely on September 21, sets up one of the largest AI infrastructure listings of the year, with the Financial Times reporting the company is seeking a valuation of up to $35 billion.
Nscale did not disclose the share count or price range. Goldman Sachs, J.P. Morgan and Morgan Stanley are lead bookrunners, with RBC, BofA, Deutsche Bank and a long list of others in supporting roles. The company plans to re-register as Nscale plc before the offering completes, converting from its current UK private structure. Founder and CEO Josh Payne stands to receive shares worth up to $350 million under a package disclosed ahead of the float, according to Business Matters.
A backlog that tripled in eight months
The prospectus gives a snapshot of how fast the AI buildout is moving. Active and contracted total contract value reached $103.4 billion as of August 31, up from $38.0 billion at the end of 2025, driven by long-term take-or-pay contracts. That is the committed revenue pipeline, not booked revenue, and the gap between the two is the main thing investors will want to probe when the roadshow starts.
| Metric | Figure |
|---|---|
| Contracted total contract value, Aug 31, 2026 | $103.4 billion |
| Contracted TCV, end of 2025 | $38.0 billion |
| Reported IPO valuation target | Up to $35 billion |
| Convertible loan package | Minimum $3.1 billion, including Nvidia participation |
| West Virginia power runway | Scalable to over 8 GW gross |
The company’s model is vertical integration. Nscale controls power generation, modular liquid-cooled data centers, GPU fleet orchestration and the cloud software layer, arguing that owning the full chain is what lets it sign take-or-pay deals at scale. The physical anchor is the Monarch Compute Campus in West Virginia, acquired through AIPCorp in March 2026, with a power generation runway scalable to more than 8 gigawatts of gross power.
Named customers include Anthropic, which signed a multi-billion-dollar infrastructure agreement, and humanoid robotics developer Figure AI, which committed to a multi-year deployment. Nscale also secured a minimum $3.1 billion convertible loan note financing package with participation from Nvidia, deepening a relationship that began with the chipmaker’s earlier investment rounds. The company raised a $900 million conventional debt round in July and has made four acquisitions along the way, folding smaller infrastructure and software teams into the platform.
The neocloud race goes public
Nscale is the third major AI cloud to head for public markets, following CoreWeave’s 2025 listing and Nebius’s Nasdaq presence. Analysts group the three as neoclouds, GPU-first infrastructure providers that rent compute to AI labs and enterprises on terms the legacy cloud vendors were slow to match. CoreWeave’s own listing set the template: a company built almost entirely on a handful of giant AI customers, valued on contracted future revenue rather than current profits.
The listing comes as AI capital spending reaches figures that would have seemed implausible two years ago. UBS now expects AI capex to approach $1 trillion this year and around $1.4 trillion by 2027. Every large lab is racing to lock in compute, and take-or-pay contracts are how they do it, which is exactly why Nscale’s backlog looks the way it does.
The same boom is straining everything around it. Intel’s CEO said this month that memory prices have risen five- to sevenfold and that his company can meet only about half of current CPU demand. SK Hynix has warned that commodity DRAM supply stays tight through 2028, and its chairman expects the broader shortage to run into 2030. Power and cooling are the next bottlenecks, which is why Nscale’s West Virginia power runway features so prominently in its pitch rather than as a footnote.
What investors will weigh
Three questions dominate. First, how much of the $103.4 billion backlog converts to recognized revenue and on what schedule, since take-or-pay commitments can be renegotiated if AI demand cools. Second, capital intensity: building 8 gigawatts of powered data center capacity requires enormous upfront spending, and the company’s cash burn between now and revenue recognition will set the dilution math. Third, competition. CoreWeave, Nebius, hyperscalers and sovereign-backed projects are all chasing the same customers, and contract values written during a shortage may look rich if supply catches up.
Counterparty concentration deserves attention too. A backlog built on a few AI labs is only as durable as those labs’ own funding cycles. Anthropic delayed its IPO to November while disclosing a revenue run rate above $100 billion, a sign the biggest customers are themselves still scaling, and OpenAI expects years of heavy cash burn ahead. If any major lab hits a funding wall, the take-or-pay contracts that underpin neocloud valuations come under renegotiation fast.
The timing is notable. AI-adjacent companies are crowding the pipeline, and Nscale will be selling into a market that has rewarded infrastructure names this year, with AMD crossing $1 trillion in market value and chip stocks rallying on the buildout, but one that has also punished any hint that AI spending might slow. The week’s mixed tape, with Anthropic pushing its listing back and Oracle cutting staff, shows how quickly sentiment swings in this corner of the market.
Pricing terms and the number of shares offered will come in subsequent filing amendments. Until then, NSCL joins the short list of pure-play ways for public investors to own the physical layer of the AI economy, alongside CoreWeave and a handful of data center REITs, with a valuation target that will test how much appetite is left for $35 billion bets on power, GPUs and contracts.