The UAE spent £18bn to own 25% of its electricity. Every hyperscaler is still renting.

The Emirates Nuclear Energy Corporation (ENEC) completed Barakah's four APR-1400 reactors in September 2024. Total programme cost: approximately £18bn ($23bn, €22bn). The UAE now owns 5.6 GW of baseload capacity outright. No Power Purchase Agreement (PPA). No renegotiation window. Twenty five percent of national electricity, secured for 60 years.
Across the Atlantic, the world's largest technology companies are taking a different path. Microsoft signed a 20-year PPA with Constellation Energy for 835 MW from the restarted Three Mile Island Unit 1. Amazon restructured its Susquehanna arrangement into a 1,920 MW front-of-meter PPA with Talen Energy. That deal is valued at £14.2bn ($18bn, €17.1bn). In the UK, Sizewell C reached financial close under a Regulated Asset Base (RAB) model with blended equity.
Three models. Three risk profiles. The structure determines who controls the energy stack for the next half century.
A PPA rents power. Equity owns the physics.

The distinction matters more than most deal teams acknowledge. A PPA is a contract for megawatt hours. The buyer pays a fixed or indexed price per unit of output. The seller retains ownership, operational control, and residual asset value. When the contract expires, the buyer renegotiates or walks away.
Equity ownership inverts the relationship. The owner bears construction risk but captures the full value of the asset across its operating life. A nuclear reactor operates for 60 to 80 years. A 20-year PPA covers a third of that lifespan at best.
The UAE understood this from the outset. ENEC holds 82% of the Barakah project. Korea Electric Power Corporation (KEPCO) retains 18% through Barakah One Company PJSC, the joint venture that manages commercial operations. Nawah Energy Company, an ENEC subsidiary, operates all four units. Sovereign capital built the plant. Sovereign entities own and operate it. The PPA question never arose because the government is both buyer and seller.
Amazon tried to plug in behind the meter. FERC pulled the cable.

Amazon's original Susquehanna deal placed a data centre behind the meter at a nuclear plant. The Federal Energy Regulatory Commission (FERC rejected the expanded interconnection agreement in November 2024, ruling that the arrangement deviated from standard grid protocols. The 480 MW behind-the-meter expansion was blocked.
The response came in June 2025. Talen and Amazon restructured the entire arrangement into a front-of-meter PPA. Susquehanna's full 1,920 MW output will flow through the PJM grid, with Talen acting as retail supplier and PPL Electric Utilities managing transmission. The transition completes in spring 2026 during a refuelling outage.
Microsoft's Constellation deal never attempted co-location. The 835 MW Crane Clean Energy Centre restart, backed by a £790m ($1bn, €950m) federal loan, delivers power to the grid. Microsoft purchases output through a 20-year PPA. Constellation retains full ownership.
In both cases, the hyperscaler rents megawatt hours. Neither owns the reactor.
If you were structuring a 1 GW nuclear deal to power an AI campus, which model would you choose?
Sizewell C costs £38bn. The UK split the bill five ways.

Sizewell C is neither a PPA nor a sovereign equity play. The RAB model collects a levy from electricity consumers during construction, lowering the developer's financing cost. The government holds 44.9%. Quebec's La Caisse holds 20%. Centrica holds 15%. EDF retains 12.5%. Amber Infrastructure takes 7.6%.
The structure spreads risk across sovereign, institutional, and ratepayer capital. Ofgem regulates returns. The RAB levy, confirmed at £3.46 per MWh, adds approximately £1.40 ($1.78, €1.68) per month to an average household bill during construction.
The trade-off is cost. Initial estimates of £12bn have escalated. Current projections sit between £36bn and £38bn ($46bn to $48bn, €43bn to €46bn). The RAB model absorbs overruns across a wider capital base. Whether that discipline holds through completion will test the model.
Susquehanna sells at £51/MWh. Crane sells at £80. Same technology. Different risk.

Pricing signals from disclosed US deals reveal the divergence. S&P Global reports three reference points: Susquehanna at £51/MWh ($65/MWh, €62/MWh), Comanche Peak at £71/MWh ($90/MWh, €86/MWh), and Crane at £80/MWh ($101/MWh, €97/MWh).
The spread reflects risk allocation. Susquehanna is an operating plant with a proven output history. Crane requires a £1.3bn ($1.6bn, €1.5bn) restart. Higher risk commands higher price. In all three cases, the operator captures the asset premium. The hyperscaler pays for certainty.
Compare the UAE. Barakah's levelised cost sits below £40/MWh ($50/MWh, €48/MWh) on most independent estimates. The sovereign owner captures both the generation margin and the avoided import cost. No intermediary takes a cut. No PPA premium applies. The same physics, the same reactor technology, produces fundamentally different economics depending on who signs the cheque.
Meta signed 6.6 GW. Google signed 500 MW. The structure matters more than the volume.

The hyperscaler nuclear wave is accelerating. Meta committed to 6.6 GW across Vistra, TerraPower, and Oklo in January 2026. Google signed a 500 MW SMR fleet agreement with Kairos Power. The deal volume is no longer the question. The structure is.
Three principles separate durable structures from fragile ones.
First, match the contract length to the asset life. A 20-year PPA on a 60-year asset leaves 40 years of uncontracted output. The residual value accrues to the owner.
Second, construction risk determines pricing. Restart deals (Crane) price higher than operating plant deals (Susquehanna) because the buyer absorbs schedule uncertainty through the PPA rate. Equity absorbs it through capital.
Third, sovereign models outperform on cost of capital. The UAE financed Barakah at sovereign rates. The UK's RAB model blends sovereign and institutional cost. US PPAs embed commercial financing costs into the per-MWh price. The cheapest megawatt hour comes from the cheapest capital.
A 20-year PPA on a 60-year asset. The owner keeps the residual.

A PPA rents the foundation of the energy stack. Equity owns it. The UAE model demonstrates what sovereign ownership looks like at scale: 5.6 GW, 25% of national supply, no intermediary, no renegotiation risk. The US model demonstrates what the market offers when private capital leads: faster deal flow, distributed risk, but structural dependence on contract renewal.
The UK's RAB experiment sits between them. If Sizewell C delivers within its revised budget, the model exports to every country with a creditworthy ratepayer base. If it overruns further, the RAB becomes a cautionary tale.
For investors evaluating nuclear exposure, the question is not whether to back nuclear power. The question is whether to rent the output or own the asset. The answer determines strategic independence for the next 60 years.
Atlas tracks all three models across 565 facilities at atlas.vistergy.com.
Next week: Three sectors. Four continents. One pattern the energy facility does not know about.
The ops hire that onboards in 30 seconds.
Viktor is an AI coworker that lives in Slack, right where your team already works.
Message Viktor like a teammate: "pull last quarter's revenue by channel," or "build a dashboard for our board meeting."
Viktor connects to your tools, does the work, and delivers the actual report, spreadsheet, or dashboard. Not a summary. The real thing.
There’s no new software to adopt and no one to train.
Most teams start with one task. Within a week, Viktor is handling half of their ops.



