LatticeLog

  • Governance
  • Infrastructure
  • Markets
  • Analysis
  • Notes
  • Signals
  • Cycles
  • Learnings
  • Commentaries
  • Glossary
  • Home

Written by Nithinraj Kooneri

in Bifrost Systems
The Nuclear Restart — Fenrir Research
Bifrost Systems/Build/The Nuclear Restart
Fenrir Research · Bifrost Systems · Build / 07

The Nuclear Restart: Signal, or Hype Cycle?

In eighteen months, nuclear went from a declining industry to the centrepiece of AI’s power strategy — roughly 10 GW committed by the hyperscalers. But restarting an old reactor and building a new one are very different bets, and only one of them arrives this decade.
Fenrir Research  ·  Jul 2026  ·  Yggdrasil Ledger / latticelog.in

It is one thing to rekindle a hearth whose stones are still standing, its chimney still sound — that is an evening’s work. It is quite another to raise a new forge from the bare hillside, and the two should never be spoken of in the same breath, though both are called making fire.

Original epigraph, in the register of Tolkien’s hearth- and forge-verses
Section 01

The Pivot Nobody Predicted

For two decades, nuclear power in the West was a story of managed decline: plants closing early, projects cancelled, an industry written off as too slow and too expensive for a renewables age. Then the AI build-out ran into the power wall — and in barely eighteen months, every major hyperscaler pivoted to nuclear at once.

The logic is the primer’s framework in action. A training cluster needs power that is firm, around-the-clock, and carbon-free — and nuclear is the only source that delivers all three at scale, running at 95%-plus capacity factors against 25–35% for solar or wind, on roughly fifty acres. Just as important, the hyperscalers brought the one thing the industry always lacked: a credit-worthy buyer willing to sign a twenty-year power-purchase agreement. Collectively they have now committed to something like 10 gigawatts of nuclear capacity across more than a dozen deals.

Committed Nuclear (Big Four)
~10 GW
Across 13+ hyperscaler deals in 2024–26
Capacity Factor
95%+
vs. 25–35% for solar / wind — the firmness premium
PPA Duration
20 yrs
Far longer than typical renewable contracts — the real unlock
New Nuclear Needed by 2030
85–90 GW
Goldman estimate; less than 10% is available globally

That last figure is the tension in miniature. The demand is real and enormous. Whether the industry can actually deliver it this decade is the entire question — and the answer depends heavily on which kind of “nuclear” a given deal really means.

Section 02

Three Very Different Bets Wearing One Label

The headlines lump everything together as “nuclear for AI,” but the deals fall into three tiers with radically different risk and timing. Confusing them is the single most common analytical error in this space.

TierWhat it isTime to powerRisk
RestartRecommissioning a recently-closed, proven reactor (Three Mile Island / Crane; Palisades)~3–5 yrsLow — known asset
Uprate / colocationSqueezing more from, or siting load at, an operating plant (Susquehanna)~2–4 yrsLow–moderate
New-build SMRPurpose-built small modular reactors (Kairos, X-energy, Natrium, Oklo)~7–10 yrsHigh — first-of-a-kind

Microsoft’s deal to restart Three Mile Island Unit 1 — now the Crane Clean Energy Center, an 835 MW reactor closed in 2019 for purely economic reasons — is a restart: proven technology, and its timeline has actually been pulled forward to 2027. Amazon’s Susquehanna arrangement is largely colocation and uprate at a running plant. But Google’s Kairos order, Amazon’s X-energy stake, Meta’s TerraPower and Oklo deals are new-build SMRs — a different universe of risk, delivering in the early-to-mid 2030s.

The Distinction That Decides Everything

Restarting a reactor is an evening’s work. Building a new one is a decade’s.

The restarts and uprates are a genuine, near-term signal: proven assets, financeable today, delivering power before 2030. The new-build SMRs are a longer-dated option — potentially transformative, but carrying first-of-a-kind cost and schedule risk that the nuclear industry has failed to control for fifty years. A portfolio that treats a 2027 restart and a 2033 SMR as the same trade is mispricing both.

Section 03

The Economics — Told Honestly

Nuclear’s cost problem has not gone away; the AI demand has simply made buyers willing to pay it. The numbers are sobering, and they are why this is a bronze-tinted story rather than a green one.

Levelised Cost of Electricity by Source ($/MWh)
Indicative LCOE ranges. First-of-a-kind SMRs are far more expensive than existing nuclear or renewables; costs are projected to fall only after 10+ GW of cumulative deployment. Sources: industry LCOE analyses (2026); IEA. Existing nuclear ~$30–60; renewables ~$20–50; FOAK SMR ~$100–180/MWh.

A first-of-a-kind SMR lands around $100–180 per MWh — several times the cost of the existing nuclear it is meant to emulate, and far above renewables. The economics only close with three things stacked together: a carbon-free mandate that rules out cheap gas, a twenty-year PPA that guarantees the revenue, and federal support — production and investment tax credits plus DOE loans. Remove any one and most new-build projects stop penciling. And the cost is a chicken-and-egg problem: SMRs only get cheap after many are built, but few will be built until they are cheap.

Analyst Read — The PPA Is the Real Innovation

The genuinely new thing here is not a reactor design — it is the twenty-year, investment-grade offtake contract. Nuclear’s historical killer was financing risk: enormous upfront cost against uncertain future power prices. A two-decade PPA from a hyperscaler with an impeccable balance sheet removes exactly that risk, which is what makes even a restart bankable. In the primer’s terms, the contract is doing more work than the physics. Watch the offtake, not the announcement.

Section 04

The Skeptic’s Case: Timelines Don’t Lie

Against the enthusiasm sits fifty years of the industry missing its own schedules. The cautionary tale is Vogtle Units 3 and 4 in Georgia — the most recent large US reactors, and “proven” AP1000 designs. They still took roughly a decade and ran past $30 billion. If proven technology behaves that way, first-of-a-kind SMRs deserve deep skepticism on both cost and schedule.

Time to Power, by Nuclear Pathway (Years)
Approximate time from decision to operation. Restarts and uprates can serve late-2020s demand; new-build SMRs and large reactors largely cannot. Sources: industry timeline analyses (2025–26). The mismatch with hyperscaler 2028 capacity needs is the core risk.

Two hard bottlenecks compound the timing problem. HALEU fuel — the higher-enriched uranium many advanced designs require — is barely produced outside Russia, and domestic supply is only now being built. And the nuclear workforce has atrophied through decades of decline, leaving a thin talent pool of licensed engineers and specialised construction crews. Money can be summoned quickly; enriched fuel and trained people cannot.

Connects to: The Power-Compute Nexus (the demand driving this) · Colocation & the Bypass Economy (the Susquehanna model) · Second-Life Infrastructure (reactors on retired coal sites) · Resource Adequacy: Power (whether any of this arrives in time).
Section 05

Signal or Hype? Both — and the Map

The honest verdict is that both readings are correct, for different tiers. The restart-and-uprate story is a real signal: proven assets, twenty-year contracts, power before 2030, and a moat in the finite set of restartable reactors. The new-build SMR story is, for now, mostly hype-adjacent optionality — potentially huge, but unproven on cost and schedule, and unlikely to matter before the 2030s. The investment map has to respect that split.

Existing Nuclear Operators
The near-term winner
Owners of operable or restartable reactors can sign 20-year PPAs today — the clearest, lowest-risk exposure to the theme.
Restart & Uprate Plays
Financeable now
Recommissioning closed plants and uprating running ones delivers this decade, backed by federal loans and credits.
SMR Developers
Option, not delivery
Real order books and milestones, but first-of-a-kind cost/schedule risk — a long-dated bet, not near-term power.
HALEU Fuel Supply
The choke point
Advanced reactors are useless without fuel; domestic enrichment is a scarce, strategically-backed bottleneck.
Nuclear Supply Chain / EPC
Capacity-constrained
Forgings, components and licensed labour are scarce — bullish for incumbents, a constraint on the whole build.
SMR Pure-Play Equities
Priced for perfection
Some valuations already discount flawless execution of unbuilt designs — the classic first-of-a-kind trap.
The Signal
Restarts and uprates are proven, financeable, and delivering before 2030
The 20-year hyperscaler PPA removes nuclear’s historical financing killer
Firm, carbon-free baseload is genuinely scarce — nuclear is the only scaled source
Restartable reactors are a finite, non-replicable set — a real moat
The Hype Risk
First-of-a-kind SMRs run $100–180/MWh and won’t deliver at scale until the 2030s
Vogtle proved even “proven” designs run a decade late and billions over
HALEU fuel and a thin talent pool are bottlenecks money can’t fix quickly
Some SMR equities already price flawless execution of unbuilt reactors
Bottom Line

The nuclear restart is real where it is boring and speculative where it is exciting. Recommissioning a proven reactor under a twenty-year hyperscaler contract is a genuine, near-term signal — the demand is enormous, the offtake is investment-grade, and the moat of restartable reactors is finite. That part of the story deserves the enthusiasm.

The new-build SMR wave is a different animal: potentially transformative, but carrying the same first-of-a-kind cost and schedule risk that has humbled the industry for half a century, and unlikely to matter this decade. Read every nuclear headline by its tier — restart, uprate, or new-build — and by its offtake. Rekindling an old fire and raising a new forge are both called making fire, but only one of them is done by nightfall.

The wise smith relit the old hearths first, for their stones were sound and their draught was true. The new forge on the bare hill he began also — but he did not warm his hands at it, nor promise its heat to anyone, until many winters had proven it would burn.

Original epigraph, in the register of Tolkien’s hearth-verses
Bifrost Systems · Build Thread
← Previous
Solar+ and Wind+
Firming variable renewables into bankable capacity
Next →
Offshore Wind: A US Post-Mortem
Why it worked in Europe and stalled in America
Sources & Notes
Deal detail & capacity: smrintel.com nuclear-for-data-centers tracker (2026); Introl and Build.inc analyses (2026); company announcements. Three Mile Island / Crane Clean Energy Center restart (Constellation–Microsoft, ~835 MW, 20-year PPA, DOE $1bn loan, targeted 2027). Amazon Susquehanna (~1.92 GW) and X-energy financing; Google–Kairos (500 MW SMR); Meta TerraPower / Oklo / Vistra commitments. Economics: industry LCOE analyses (existing nuclear ~$30–60/MWh, FOAK SMR ~$100–180/MWh, renewables ~$20–50/MWh); nuclear ~$6,400–12,700/kW vs. gas ~$1,290/kW; Goldman Sachs (85–90 GW new nuclear needed by 2030). Vogtle 3&4 cost/schedule; HALEU and workforce constraints per DOE and industry reporting. Figures are indicative and the most recent available as of publication. All framing and conclusions are Fenrir Research’s own.
This analysis is for informational purposes only. Not investment advice. Company and project references are illustrative of sector dynamics, not recommendations. Fenrir Research is a division of Yggdrasil Ledger (latticelog.in).
←Solar+ and Wind+
Offshore Wind: A US Post-Mortem→

Comments

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

More posts

  • The Health Case That Closes

    July 28, 2026
  • The Border Adjustment Problem

    July 28, 2026
  • The Young Fleet

    July 28, 2026
  • Committed Emissions

    July 28, 2026

LatticeLog

Structural research across markets, infrastructure, climate, and the systems that connect them. Published under Fenrir Research, a division of Yggdrasil Ledger.

  • Blog
  • About
  • FAQs
  • Authors

Twenty Twenty-Five

Designed with WordPress