99% sovereign capital. 1% commercial bank.

Abu Dhabi financed Barakah without a single commercial bank lead. The £19bn ($24.4bn, €23.3bn) project drew £12.6bn ($16.2bn, €15.5bn) from the Abu Dhabi Department of Finance. The Export-Import Bank of Korea contributed £1.9bn ($2.5bn, €2.4bn). Commercial banks provided £195m ($250m, €239m). That is 1% of total financing. The other 99% came from sovereign capital.

Capital markets across Europe and North America paused for Easter this week. Sovereign infrastructure programmes did not. The financing models that built Barakah operate on a different calendar.

Last week, we introduced the five-layer infrastructure stack. Energy at the base. GDP at the top. Five dependencies between them. This week, we apply it to financing. The result exposes a structural mismatch: banks price Layer 1 as commodity risk. Sovereign wealth funds price it as strategic infrastructure. The gap between those two prices explains why nuclear for AI remains unfunded at scale.

Traditional project finance prices nuclear against wholesale electricity markets. The IEA notes that cost overruns, long construction timelines, and volatile power prices make banks reluctant to lead nuclear financing. A bank sees a power plant competing with gas on levelised cost of electricity (LCOE). The loan committee models merchant risk: what happens if electricity prices fall?

But nuclear for AI is not a merchant asset. A hyperscaler signing a 20-year power purchase agreement (PPA) does not care about LCOE in year 15. It cares about Layer 1 availability in year 2. The buyer is not purchasing electricity. The buyer is purchasing infrastructure certainty.

This is the mismatch. Banks have no model for "strategic Layer 1 capacity secured against AI demand." They model "electricity generation exposed to wholesale price risk." These are fundamentally different products. The credit committees are pricing the wrong layer.

UAE: Sovereign capital bypasses banks entirely. Barakah's financing structure concentrated 99% of capital in sovereign and export credit sources. The Abu Dhabi model treats nuclear as a national infrastructure asset, not a commercial energy investment. KEPCO holds 18% equity through Barakah One PJSC. The result: 5.6 GW operational. No refinancing risk. No market exposure.

France: State de-risks, then invites capital. EDF's six EPR2 reactors carry an estimated £61bn ($72.8bn, €69bn) price tag. France structured a subsidised loan from Caisse des Dépôts covering 60% of costs, paired with a 40-year contract for difference (CfD). The state absorbs construction risk. Private capital prices operational risk. Final investment decision targets the end of 2026.

UK: Consumer levy during construction. Sizewell C reached financial close in November 2025 using the Regulated Asset Base (RAB) model. £5bn in export credit from BpifranceAE. Total capital: £8.8bn equity plus £41.6bn debt. The UK government holds 44.9%. La Caisse holds 20%. Consumers pay a levy of 0.36p/kWh from December 2025. The RAB model spreads risk across sovereign, institutional, and consumer capital.

United States: No sovereign model. Hyperscalers fill the gap. Constellation signed a 20-year PPA with Microsoft for Three Mile Island's 835 MW restart. The DOE provided a £780m ($1bn, €953m) loan in November 2025. Amazon signed a 17-year PPA with Talen Energy for 1.92 GW from Susquehanna. Corporate PPAs are substituting for sovereign financing. The hyperscaler becomes the de-risking entity that a sovereign wealth fund would be elsewhere.

The National Center for Energy Analytics reports £264bn ($335bn, €321bn) of infrastructure dry powder sitting undeployed as of December 2024. Infrastructure assets under management grew from £77bn ($98bn, €94bn) in 2007 to £1.1tn ($1.4tn, €1.3tn) in 2024. The capital exists. It is not flowing to nuclear.

The reason is layer mismatch. Infrastructure funds price nuclear as a Layer 1 commodity asset: generation competing with renewables and gas. Sovereign wealth funds price nuclear as a Layer 2 strategic asset: infrastructure enabling AI, defence, and industrial competitiveness. The same reactor, priced in two different layers, produces two different risk profiles and two different returns.

KEPCO proposed building APR1400 reactors on American soil in February 2026, part of a £157bn ($200bn, €191bn) investment pledge. South Korea targets 10 reactor exports by 2030. The financing model is inseparable from the technology: Korean export credit, KEXIM backing, and sovereign guarantees travel with the reactor. The buyer receives a Layer 1 asset and a Layer 1 financing package.

This contrasts with the US approach, where the reactor exists but the financing model does not. Constellation has the asset. Microsoft has the demand. But no sovereign entity de-risks the construction phase. The DOE's £780m ($1bn, €953m) loan for TMI is a fraction of Barakah's £12.6bn sovereign commitment. Scale requires sovereign conviction.

First, corporate PPAs will evolve into equity positions. Microsoft's 20-year PPA with Constellation is a financing instrument disguised as an energy contract. The next step is direct equity: hyperscalers taking ownership stakes in nuclear assets. When the buyer becomes the owner, Layer 1 and Layer 3 integrate financially, not just operationally.

Second, sovereign wealth funds will enter nuclear directly. Abu Dhabi proved the model. Saudi Arabia's Nuclear Holding Company, established in 2022, signals intent. The PIF targets £2.1tn ($2.67tn, €2.6tn) in assets by 2030. Nuclear fits a sovereign portfolio better than a bank balance sheet.

Third, the RAB model will spread. Sizewell C demonstrated that consumers, sovereigns, and institutional investors can share construction risk. India, Poland, and the Czech Republic are watching. The model de-risks Layer 1 for Layer 3 buyers. Expect replication.

The five-layer model reveals why nuclear financing is stuck. Banks price Layer 1 against other Layer 1 assets: gas, wind, solar. The return profile is commodity-grade. Sovereign funds price Layer 1 against Layer 5 outcomes: GDP growth, AI capability, energy independence. The return profile is strategic-grade.

The £264bn ($335bn, €321bn) of undeployed infrastructure capital is not waiting for better nuclear technology. It is waiting for a pricing model that matches the asset to the right layer. The organisations that see nuclear as AI infrastructure, not energy infrastructure, will finance the next decade.

The question is not whether nuclear can be financed. Barakah, Sizewell C, and the French EPR2 programme prove it can. The question is whether your financing model prices the right layer.

Sovereign investment committees decide. The model informs their conviction.

Next week: 50% faster to power. What our partnership means.

You don't have to read everything — just the right thing. 1440's daily newsletter distills the day's biggest stories from 100+ sources into one quick, 5-minute read. It's the fastest way to stay sharp, sound informed, and actually understand what's happening in the world. Join 4.5 million readers who start their day the smart way.

Join for free today!