Rolls-Royce's 'Imminent' Hyperscaler Deal Pulls a UK Nuclear Supplier Into the US AI Power Trade

Erginbilgic's exact phrasing — \"Early next week, in [the] investment committee, we are going to sign another big hyperscaler sort of deal\" — is a deliberate signal. It tells the market (1) the customer has been identified, (2) commercial terms are agreed, (3) only internal capital approval remains, and (4) this is one of several, not a one-off. CFO Helen McCabe separately disclosed that data-center orders across the Power Systems business grew more than 50% in H1 2026, and that the order book already extends into 2028.

That sequencing — gas engines now, SMRs after 2030 — is why the deal is described as \"hyperscaler nuclear\" rather than \"hyperscaler SMR\". Rolls-Royce's installed mtu Series 4000 fleet can be commissioned in months, not years. Until an SMR reaches commercial operation, the cloud customer is buying optionality on nuclear-grade baseload plus immediate relief from grid interconnection queues. Rolls-Royce Power Systems FY2025 order intake rose 21% to €7.14B, with original-equipment order coverage at 100% for the remainder of 2025 and 43% for 2026 even before this week's announcement.

Rolls-Royce SMR's supply chain is the most concrete read on which US-listed companies get a second-order lift. The reactor vessel, fuel and components run through BWX Technologies, the only US-domiciled manufacturer with naval and commercial nuclear credentials; fuel enrichment runs through Centrus Energy (the only HALEU producer in the US, which received a $900M DOE order in January 2026 to expand centrifuge cascades); instrumentation and control systems now run through Yokogawa Electric of Japan; and uranium enrichment more broadly is being expanded by Urenco USA, which announced a 2.1 million SWU capacity expansion in New Mexico on June 2, 2026.

The transmission into US-listed equities is uneven. BWXT sits in the most direct slot — even before any Rolls-Royce SMR order, BWXT's mPower design was selected by Core Power for floating nuclear plants in June 2026, and the company is positioning its BANR reactor for the same hyperscaler market. Centrus Energy is the bottleneck input: every SMR design eventually needs fuel, and the US is structurally short of enrichment capacity, which is why Centrus trades at 35x 2026 EV/EBITDA on the bull-case math despite quarterly earnings volatility.

No customer has been named, but the hyperscaler-of-nuclear-bets map narrows the field. Microsoft is locked into Constellation Energy via the 20-year Three Mile Island PPA (priced by Jefferies at $110–115/MWh). Amazon bought Talen's 960MW Susquehanna-powered campus for $650M in 2024, signed a deal with Dominion Energy for small modular reactor development, and announced plans for up to 12 SMRs. Alphabet signed the world's first corporate SMR offtake with Kairos Power in October 2024 and is developing multiple follow-on agreements. Meta signed across four nuclear partners in late 2025 for a combined ~7.7 GW. Oracle, Apple, and Microsoft's smaller commitments leave a thin remaining universe of uncommitted hyperscalers — most plausibly an EMEA-focused cloud operator (Deutsche Telekom / T-Systems, BT Group, Orange) or an Asia-Pacific hyperscaler (Tata, NTT).

The valuation lens matters. Rolls-Royce is the first non-US industrial to convert AI-power demand into a quantitative guidance raise — every US merchant generator re-rating since 2024 has been about the same thesis, but Vistra, Talen and Constellation are all rated for in-service nuclear capacity they don't own. Rolls-Royce owns the design, the manufacturing blueprint, and the operating IP. At a current £122B market cap on 28.3x trailing EV/FCF, the market is paying for execution, not narrative; the hyperscaler deal merely confirms the execution is real.

The non-obvious read: a successful hyperscaler deal is more valuable for the supply-chain pull-through than for Rolls-Royce's own P&L. At £1.8B per unit and 470MW output, even a 5-unit hyperscaler fleet adds ~£9B of revenue spread over a decade — meaningful, but only ~7% of cumulative group revenue. The bigger number is the implied downstream build: a hyperscaler offtake typically anchors a 5–10 unit pipeline, and each unit pulls through ~$200–400M of components, instrumentation and fuel purchases that flow to BWXT, Centrus, and Yokogawa.

The thesis breaks if any of three things happen: a US-domiciled SMR designer (most plausibly TerraPower with Microsoft, or X-energy with Amazon) wins a hyperscaler RFP that was previously in Rolls-Royce's pipeline; the UK planning system blocks Wylfa past 2027 FID; or hyperscaler capex pauses materially as AI training infrastructure matures. None of these is the base case — Rolls-Royce is the only SMR developer with multiple contracted fleets, a control-systems partner installed, and a CEO publicly pricing the order book out to 2028.