A $103 Billion Backlog, Two Clients, and a $35 Billion Valuation Question
TREE NEWS reports: British AI data center developer Nscale is seeking a US listing, and its S-1 filing reveals a striking concentration risk: roughly 85% of its $103 billion contract value comes from just two customers — Microsoft (about $43.8 billion) and Anthropic (about $44.6 billion). The company is targeting a valuation of up to $35 billion, even as its active contracts stood at only $2.6 billion at the end of August.
The gap between headline backlog and realized revenue is the central issue. In the first half of this year, Nscale generated $140.6 million in revenue while posting a net loss of $1.02 billion. More critically, the $44.6 billion Anthropic agreement — tied to an 8-gigawatt data center campus in West Virginia, with the first 2 GW expected online in 2028 — carries no binding financing commitment. It is subject to milestone requirements and strict continuous compute-supply standards, meaning it could be terminated if Nscale fails to deliver.
Why the Market Should Care
This is not just an IPO story. It is a stress test of the entire AI infrastructure trade, which has driven a meaningful share of US equity gains over the past two years. Nscale sits at the intersection of three heavily correlated exposures: hyperscaler capex, AI chip supply chains, and private credit-financed data center construction.
- Equities: A successful, well-subscribed IPO would validate demand for AI infrastructure names and likely lift peers such as CoreWeave, Nebius, and other neocloud operators. A weak reception — or a valuation cut — would force a repricing of the sector’s growth assumptions and could spill over into semiconductor stocks, particularly Nvidia, which is both a shareholder in Nscale and a supplier, and which recently participated in a $3.1 billion financing round including $1 billion in convertible notes or non-voting shares.
- Credit and rates: The AI buildout is increasingly funded through debt and leasing obligations. Nvidia supports roughly $860 million of Nscale’s lease obligations. If investors begin to question whether forward contracts convert into cash flow, spreads on data center ABS and private credit could widen, tightening financial conditions for the whole sector.
- Crypto: AI-adjacent crypto tokens and decentralized compute networks have traded as a leveraged proxy for the AI capex cycle. A credibility shock at a flagship AI infrastructure IPO would likely hit those tokens harder than traditional equities, given their thinner liquidity and narrative dependence.
- Commodities and FX: Data center construction is power-intensive and hardware-heavy. Any slowdown in confirmed buildouts would soften demand expectations for electricity, copper, and cooling infrastructure, while a stronger dollar narrative — if US AI leadership is reaffirmed — could weigh on GBP given Nscale’s UK roots.
Customer Concentration Is an Industry Feature, Not a Bug
Nscale is not an outlier. Microsoft and Meta together accounted for roughly half of Nebius’s annual revenue, and Microsoft represented about 67% of CoreWeave’s 2025 revenue. The pattern reflects the reality that only a handful of hyperscalers can underwrite gigawatt-scale projects. But it also means that any shift in a single customer’s capex plans — or a contract cancellation, as happened when OpenAI exited a joint Stargate project in Norway and the UK, with Microsoft and Google stepping in — can cascade through an entire company’s valuation.
Key Takeaways for Investors
- Backlog is not revenue. The $103 billion figure is a pipeline, not cash. Only $2.6 billion is active. Watch the conversion rate, not the headline number.
- Financing is the binding constraint. Without a binding commitment for the Anthropic project, the largest slice of Nscale’s backlog is contingent. This is a financing story as much as a technology story.
- Nvidia’s dual role cuts both ways. Equity ownership and supply support reduce near-term risk but do not eliminate delivery delays on next-generation accelerators.
- Watch the IPO pricing. If the deal prices at or below the $35 billion target with strong demand, it signals confidence in AI infrastructure. If it prices down or struggles, expect a broader reassessment of AI-linked equities, credit, and crypto compute tokens.
