Nuclear construction in America doesn't have a cost problem. It has an architecture problem.

Every project is treated as a one-off. No standardized financing. No shared learning. No credible cost cap. The result: unbounded risk that no one can price and no one will finance.

See the solution ↓
The U.S. built 100+ reactors. Costs went up, not down. Learning rate: −115%

What NCFC Is

NCFC is a proposed federally chartered institution that standardizes, pools, and securitizes NOAK nuclear construction debt. Think Freddie Mac for nuclear construction — it doesn't build plants, it makes building them financeable.

The core insight: capital providers will finance nuclear if someone credibly bounds the tail risk. NCFC does this through a structured commercial loss waterfall, a deep-tail Completion Reserve Facility, milestone-based take-outs that tie sovereign exposure to verified physical progress, and a master trust that converts bespoke construction loans into a standardized, tradeable asset class.

The sovereign bargain: the federal government is already backstopping nuclear cost overruns. LPO funded Vogtle's $35 billion finish. Congress absorbed V.C. Summer's $9 billion failure. NCFC doesn't create new sovereign exposure — it formalizes what already happens ad hoc, makes it rules-based and bounded, and in exchange, the entire fleet conforms to the institutional discipline that produces learning curves. This is not a metaphor. The United States built over 100 reactors between 1967 and 1990 and costs rose with every doubling of experience — a measured learning rate of negative 115%. France and Korea, building under centralized institutional frameworks with standardized designs, unified procurement, and continuous sequencing, achieved strong positive learning curves over the same period. The difference is not the reactor. It is the system surrounding it.

Standardized Financing

Every NOAK project financed under the same templates, the same milestones, the same rules.

Pooled Risk

Construction notes aggregated into a master trust. Diversification replaces bespoke exposure.

Sovereign Backstop

Deep-tail completion risk owned by the sovereign — bounded, priced, and structured.

Learning Curve

The U.S. built 100 reactors without a learning curve. France and Korea built fewer — and got one. The difference: institutional architecture. NCFC builds that architecture.

The Sovereign Stack

Nuclear Deployment Council (NDC)

Sets strategy, picks NOAK designs, defines milestones

Nuclear Construction Finance Corporation (NCFC)

Finances NOAK construction, manages tail risk, creates the asset class

DOE Loan Programs Office (LPO)

Finances FOAK and frontier projects, political one-offs

Federal Demand Frameworks

PPAs, CfDs, capacity contracts anchor revenue

The line is simple: if NDC certifies it as NOAK, it goes to NCFC. If it's FOAK or frontier, it stays with LPO. No waivers. No political exceptions inside NCFC.

NDC and NCFC sit side by side — NDC defines the playing field, NCFC finances within it. No interlocking boards. Coordination through data flows, not shared decision rights. NDC contingency provisions ensure NCFC can operate even if NDC is partially staffed or absent, but at reduced capacity.

Why this matters empirically: MIT's large-scale review of U.S. nuclear construction found that 72% of total cost escalation was driven by soft costs — engineering services, field supervision, project management, and QA — not by hardware or materials. These are precisely the cost categories that institutional architecture controls. Standardized cost codes, unified baselines, supplier qualification, and cross-project data integration address the actual cost drivers. NCFC's institutional mandates are not bureaucratic overhead. They are the mechanism that compresses the fat tail.

From Bank Loans to Nuclear Construction Bonds

1 Banks originate construction loans

For NDC-qualified NOAK projects under standardized NCFC templates.

2 Physical milestones reached

Independent engineer certifies verified physical progress.

3 NCFC purchases ~80% of funded exposure

At each milestone gate.

Banks retain ~20% strip through COD (skin in the game)

4 Standardized Construction Notes

Purchased slices become fixed coupon, senior secured, first-lien notes.

5 Notes pool into the NCB Master Trust

Diversified portfolio of standardized construction debt.

6 Trust issues Nuclear Construction Bonds (NCBs)
  • Class A Senior (~80%): Federally guaranteed
  • Class B Mezzanine (~15%): Unguaranteed, rated on pool risk
  • Residual (~5%): NCFC/CRF retained first-loss

The milestone take-out structure means NCFC's exposure is always tied to verified physical progress — not promises, not schedules, not percent-of-budget. Banks keep real skin in the game through COD. The master trust creates a liquid, standardized asset class from what would otherwise be bespoke, illiquid construction debt.

The Loss Architecture

Before any sovereign capital is deployed, a substantial commercial loss stack absorbs mid-band cost overruns. This stack was validated through engagement with infrastructure financing professionals at major investment banks. The architecture is continuous — there is no gap between the commercial stack and the sovereign backstop.

Dynamic Hyperscaler PPAs ~20-30% absorption
Equity IRR Erosion 16% → 10% target
OEM Partial Wraps Component guarantees
EPC Partial Wraps Lump-sum/GMP scope
Technology Sponsor Slices Strategic investor risk
Commercial Stack Exhausted
CLT — Completion Loan Tranche ~160% of B₀
CTT — Catastrophic Tail Tranche ~180% of B₀
Hard Tail Cap: ~1.8–2.0× original budget

The hard tail cap is the critical signal to capital providers: "Whatever happens, this project will not cost more than X% over budget." That bounded risk is what makes nuclear construction financeable. Without it, the right tail is infinite, and infinite risk is uninsurable and unfinanceable.

Structured, Not Ad Hoc

Nuclear cost overrun risk is not insurable. The historic U.S. cost distribution has a median of 1.67× — meaning the typical project finishes 67% over budget. No private insurer can underwrite a risk where the majority of outcomes exceed the policy limit. The federal government is already absorbing this tail. The question is whether it does so by accident — or by design.

Status Quo Under NCFC
Median U.S. nuclear project finishes 67% over budget (Budzier & Flyvbjerg, 2018) Structural standards compress the cost distribution — variance reduction + median improvement
72% of cost escalation is soft costs no one tracks consistently (MIT) Standardized cost codes, unified baselines, mandatory reporting target the actual cost drivers
U.S. built 100+ reactors: learning rate of −115% (Lovering et al., 2016) Institutional architecture that produces learning curves: design freeze, supplier continuity, fleet sequencing
V.C. Summer: $9B spent, zero electricity. Vogtle: ~$35B for 2.2 GW Hard tail cap at ~1.8-2.0× budget. Viability gates. Orderly wind-down before catastrophe
Each project is de facto first-of-a-kind — no shared data, no fleet learning Each project contributes to the data spine. China's second AP1000 series: construction durations cut by ~30-40% at major milestones
Sovereign exposure is ad hoc, unlimited, and politically negotiated Sovereign exposure is bounded, rules-based, and structurally committed

What industry gets: Access to a sovereign completion platform that bounds tail risk and makes nuclear construction financeable at scale. The hard tail cap at ~1.8-2.0× budget is the signal capital providers need: whatever happens, exposure is bounded.

What industry gives: Full conformity to the institutional discipline package — standardized cost codes, unified reporting, NDC milestone frameworks, mandatory post-mortems, data spine contributions — that compresses the learning curve. This is not optional. This is the price of admission.

What the sovereign gets: A fleet that actually learns. The empirical record is unambiguous: the U.S. built over 100 reactors without a learning curve because the system couldn't propagate learning. France and Korea achieved strong learning curves because their systems were designed to. NCFC builds that system. CRF premiums are the alignment capital that makes participation non-free. The data discipline is the mechanism that shifts the cost distribution left with each successive build.

CRF premiums are alignment capital — the price of admission to the sovereign completion platform — not actuarial insurance premiums. Nuclear cost overrun risk is not insurable through traditional pricing. The sovereign accepts deep-tail completion risk as a permanent structural commitment. This is not a temporary startup condition to be patched. It is a deliberate design choice.

LTRF: The Emergency Lane That Closes Naturally

LTRF exists to prevent economically viable nuclear projects from being forced to refinance into stressed capital markets at the wrong moment. It is not a subsidized long-term mortgage. It is not meant to be cheap in normal markets. It is a sovereign emergency lane that closes naturally in normal traffic.

Zone A

Good Markets

LTRF is expensive relative to market. Rational behavior: refinance immediately at COD. LTRF unused or used briefly.

Zone B

Normal Markets

LTRF approximately fair at COD, but step-ups make it expensive within 2-3 years. Rational behavior: refinance at first reset window.

Zone C

Stressed Markets

Market spreads blow out. Even after step-ups, LTRF is cheaper than distressed issuance. Developers stay 3-7+ years. Refinance when markets normalize.

Step-up pricing (+25-37.5 bps/year after a 24-month lock) creates a ratchet that naturally pushes sponsors off LTRF in normal conditions. The structural seniority of LTRF means sponsors can't form a normal capital structure until they refinance NCFC out — this blocks permanent dependence without explicit penalties.

NCFC Canonical Architecture v3

The complete NCFC Canonical Architecture — 20 sections covering every component of the system. Click any section to expand. Download the full document below.

Download the Architecture

Download the complete NCFC Canonical Architecture as a Word document.

Download .docx

Request the Adversarial Audit

We stress-tested every component of this architecture. The adversarial audit contains the full attack/response analysis — what holds, what needed strengthening, and what remains genuinely unresolved.

Start a Conversation

If you work in nuclear policy, infrastructure finance, or energy deployment and want to discuss this architecture, reach out.

Contact

Arthur Spoerl — March 2026