Key Takeaways

  • Nscale reports its contracts represent about $103 billion in total revenue over an average term of 5.7 years.
  • The portfolio implies roughly $18 billion in annualized contracted revenue, although the figures are illustrative rather than formal guidance.
  • A reported agreement with Anthropic worth tens of billions of dollars highlights both the strength of AI computing demand and Nscale's exposure to large customer commitments.

Nscale is presenting prospective investors with a striking measure of future demand: approximately $103 billion in total contracted revenue as the London-based AI cloud infrastructure provider considers a possible U.S. initial public offering.

The contracts have an average duration of 5.7 years and imply roughly $18 billion in annualized contracted revenue, according to Reuters. Another source cited in the report cautioned that the figures are illustrative and should not be treated as formal revenue guidance.

That distinction matters. Contracted revenue can offer visibility into future business, but it is not the same as revenue already recognized under accounting rules. Delivery schedules, infrastructure availability, customer usage, contractual conditions, and construction milestones can all influence when, or whether, the full amount reaches Nscale's financial statements.

Despite these caveats, the sheer size of the portfolio highlights massive infrastructure demand.

Reuters reported that Anthropic committed tens of billions of dollars over six years to rent AI cloud computing capacity from Nscale. This commitment illustrates how developers of large AI models are reserving access to power, Nvidia GPUs, and data center capacity years in advance rather than relying entirely on computing resources available in the spot market.

AI cloud contracts at this scale act as both technology agreements and infrastructure financing mechanisms. Long commitments help Nscale demonstrate demand when arranging capital for data center construction, energy infrastructure, and server purchases. Customers, meanwhile, gain predictable access to scarce computing capacity.

The market backdrop helps explain the urgency. IDC forecasts worldwide AI infrastructure spending will reach $487 billion in 2026, up from $318 billion in 2025, and exceed $1 trillion by 2029 (source). Servers with embedded GPUs generated a record $125.3 billion in revenue during the fourth quarter of 2025. GPU servers represented more than half of the server market and contributed to $444.1 billion in full-year 2025 server revenue.

Separately, Grand View Research valued the global AI data center market at $147.3 billion in 2025 and projects it will increase to $180.6 billion in 2026. Its longer-range forecast puts the market at $810.6 billion by 2033, reflecting a 23.9% compound annual growth rate.

For investors, however, demand represents only one side of the equation. The operational challenge lies in whether Nscale can deploy enough infrastructure on schedule while managing financing and operating costs.

AI data centers require substantial capital before customer revenue begins arriving. Nscale needs access to GPUs, electrical equipment, cooling systems, network capacity, and suitably powered sites. Delays in utility interconnections or hardware delivery could shift revenue recognition even when the underlying customer contract remains intact. Nvidia's product supply and roadmap are therefore relevant to Nscale's execution, while Microsoft and other hyperscale cloud operators remain important competitive reference points.

Customer concentration also requires attention. An Anthropic commitment worth tens of billions of dollars accounts for a significant portion of the $103 billion contract total. A large anchor customer supports financing and utilization, but dependence on a small number of AI model developers may introduce renewal, credit, and negotiation risks.

Security and operational controls will become another test as Nscale expands. Enterprise and model-provider workloads can involve sensitive code, proprietary training data, and valuable model weights. Controls based on NIST SP 800-53 Rev. 5, 2020 and an information security management system aligned with ISO/IEC 27001:2022 can help establish how access, incident response, supplier risk, and data center operations are governed.

A potential U.S. IPO would give public investors a chance to evaluate whether Nscale can convert an unusually large contract book into durable cash flow. The $103 billion figure establishes ambition and demand visibility. The harder work comes next: financing construction, delivering capacity, and turning multiyear promises into recognized revenue without letting costs outrun growth.