Nscale Claims $103B in Contracted Revenue Before Its September IPO

Nscale, a U.K.-based AI infrastructure startup backed by Nvidia, is telling potential investors it holds roughly $103 billion in total contracted revenue, according to documents reviewed by The Information and cited by Reuters. The company may launch an IPO as early as September 2026. The headline figure carries a significant catch: those contracts run an average of 5.7 years, putting annualised contract income at around $18 billion. One person close to the negotiations told Reuters the numbers were "illustrative" and not formal revenue guidance.

What happened

Detail	Figure
Total contracted revenue	$103 billion
Average contract length	5.7 years
Implied annualised contract income	~$18 billion
Anthropic commitment (West Virginia data center)	$45 billion
Anthropic share of total contracted figure	~44%
Planned IPO timing	As early as September 2026
Processors planned for Anthropic workloads	Nvidia Vera Rubin

Nscale is one of a new wave of specialist cloud providers building the compute infrastructure that AI labs need to train and run large models. Rather than competing with Amazon, Google, or Microsoft directly, these firms raise capital, sign long-term compute leases with AI developers, and deploy the resulting cash almost entirely into Nvidia hardware, power, and data-center capacity.

The $103 billion figure is contracted future revenue, not money already earned. That distinction matters: Nscale still needs to install the processors, networking, cooling, and physical data-center space to actually deliver what clients have agreed to buy.

Why the Anthropic deal is the key data point

Reuters reported in late August that Anthropic, the AI safety company behind the Claude models, will pay $45 billion to lease capacity from Nscale’s West Virginia facility. That single commitment accounts for nearly 44% of the $103 billion total. Nscale will use Nvidia’s Vera Rubin processors, the company’s next-generation chips, to handle Anthropic’s workloads.

The deal illustrates a broader shift in how frontier AI labs source compute. Rather than building their own data centers from scratch, labs are signing multiyear contracts with infrastructure specialists who absorb the capital risk and complexity. For Nscale, long-term deals with creditworthy clients like Anthropic are exactly what a company needs before going public: evidence that demand is real and sticky.

For Nvidia investors, firms like Nscale represent a secondary demand signal. The chipmaker’s biggest customers are large tech platforms, but infrastructure specialists locking in multiyear contracts around Vera Rubin processors extend the visibility of future chip demand well beyond Nvidia’s own quarterly reports. You can follow broader coverage of this infrastructure trend in our analysis of AI inference as the new compute bottleneck.

Why it matters

If Nscale does go public in September, it would be one of the first pure-play AI infrastructure companies to list. Public market investors would be betting on two things at once: that AI compute demand stays strong, and that Nscale can convert a large stack of long-term promises into recognised revenue while building out enough capacity to honour them.

AI infrastructure is capital-intensive before a single dollar of revenue hits an income statement. Processors, energy contracts, networking equipment, cooling systems, and real estate all need to be in place first. That execution risk is real, and it is separate from the demand story.

The “illustrative” label on the figures also introduces uncertainty. Nscale has not issued formal revenue guidance, so the $103 billion is best understood as a description of the contract pipeline, not a financial forecast.

Our take

The $103 billion number is striking, but the more interesting figure is the $18 billion annualised run-rate. That is still an enormous number for a startup, and the Anthropic anchor tenant gives it credibility. The caveat about figures being “illustrative” is the kind of language that, in a public filing, would attract close scrutiny from analysts.

For businesses thinking about their own AI compute costs, the dynamic here is worth watching. As specialist providers like Nscale compete for AI lab contracts, the alternative for smaller buyers is renting capacity from the same hyperscalers that the big labs are increasingly trying to bypass. That competitive pressure could eventually affect pricing down the stack. If you are evaluating how AI integration fits into your own infrastructure budget, the direction of travel in compute pricing is a real input to that decision.

Whether Nscale’s IPO lands in September or slips, the company’s contract stack tells you something concrete: AI labs are willing to sign decade-scale compute agreements right now, and they are paying prices that make $103 billion totals possible. That is a useful reality check against anyone arguing that AI investment is about to slow down.

What to do about it

1. Watch the Nscale IPO filing for formal revenue breakdowns. The S-1 or F-1 document will clarify how much of the $103B is truly contracted versus conditional.
2. Track Nvidia Vera Rubin availability. Multiyear deals around this chip suggest it will be the dominant training and inference processor for the next product cycle, which affects which cloud regions and services you can access.
3. If you use AI APIs (OpenAI, Anthropic, etc.), note that your vendor’s compute costs are increasingly locked into long-term agreements like this one. Price stability at the API level may improve as a result.
4. If you are exploring workflow automation that relies on AI APIs, factor compute supply into your vendor selection, not just current pricing.

The headline is big, but the execution challenge is bigger. Watch the IPO documents for the real numbers.

Source: Bing News · Anthropic