The Atomic
Enterprise

1.1The global construction map

Roughly 80 reactors are actively under construction worldwide — about 81,000 MW of potential net capacity, spread across 15–17 countries. The build cycle is heavily concentrated in Asia, and a single country drives nearly half of it.

• China's surge. Expanding faster than any country in history; once its pipeline finishes, its capacity will eclipse France and trail only the U.S.
• Russia's reach. Only 5 reactors at home, but Rosatom is funding, building, or supplying tech for 20+ reactors globally — the real export superpower.
• Western pause. After Vogtle Unit 4 (U.S.) and Flamanville 3 (France), large Western construction largely stopped — leaving Hinkley Point C and a few SMR pilots.

1.2The American picture: pause, then reversal

After Plant Vogtle Unit 4, the U.S. has zero traditional large light-water reactors under construction. But "no reactors being built" is wrong — the answer split into two tracks at once.

The common driver is the AI energy crunch: wind and solar are too intermittent to run 24/7 AI clusters, so tech giants are chasing firm, grid-independent power.

1.3Small modular & advanced reactors

A genuine domestic renaissance is underway in next-generation designs — several have broken ground or are fueling.

• TerraPower "Natrium" — Kemmerer, WY. 345 MW sodium-cooled fast reactor with molten-salt storage. Co-founded by Bill Gates; NRC construction permit granted March 2026. (Full case study below.)
• Kairos Power "Hermes 2" — Oak Ridge, TN. 50 MW molten-salt-cooled demonstrator, under a power-purchase deal with TVA to feed Google data centers.
• Antares "Mark-0" & Radiant "Kaleidos" — Idaho National Lab. Trailer-sized microreactors (~1 MW class) tested at INL's DOME facility; aimed at military bases and remote industry, with field deployment targeted around 2028.

1.4The return of the gigawatt giants

The energy crunch forced an about-face. The strategy is no longer "SMRs instead of large reactors" but both at once.

• Federal mandate. An executive order ("Reinvigorating the Nuclear Industrial Base") targets 300 GW of new nuclear by 2050, with a near-term goal of 10 large reactors under construction by 2030, prioritized through the DOE Loan Programs Office.
• The $80B fleet partnership. Westinghouse (maker of the 1,117 MW AP1000 at Vogtle) formed an $80B-backed effort to build large reactors using a "fleet-scale" copy-paste model — submitting Vogtle Unit 4 as the official "as-built reference plant" so utilities skip years of custom review. Suppliers like Cameco see a path to ~20 new AP1000s.
• Reviving dead projects. The prime candidate is V.C. Summer (South Carolina), where two AP1000s were abandoned in 2017 after billions spent — now targeted for a federally backed restart.

1.5Case study — TerraPower Natrium

At 345 MW, Natrium straddles the line between "small modular" and gigawatt-scale (a third of an AP1000). It's the flagship test of whether advanced reactors can actually be built on time and on budget.

The "separation" build advantage

Vogtle's delays came from everything inside the plant needing strict NQA-1 nuclear-quality paperwork. TerraPower split the plant in two so only a tiny footprint carries that burden:

The real bottleneck isn't construction — it's fuel. Natrium needs HALEU (high-assay low-enriched uranium, up to 20%). Russia was the only commercial supplier; the Ukraine war pushed the launch from 2028 to 2030 while U.S. suppliers (Centrus) ramped domestic HALEU. Meta has signed on to fund up to eight Natrium units.

2.1How a naval reactor works

They are Pressurized Water Reactors (PWRs) — the same primary/secondary-loop concept as a commercial plant. In fact the entire commercial industry descends from Rickover's 1950s naval program (the first commercial U.S. plant, Shippingport, was a scaled-up sub reactor). On a ship, the steam does double duty: it spins reduction gears to turn the propeller shafts and drives turbogenerators for electronics, weapons, and (on carriers) the electromagnetic catapults.

Four ways naval PWRs differ from commercial ones

• Highly enriched fuel. Commercial uses 3–5% U-235; U.S. naval reactors use >93% HEU — essentially weapons-grade. This makes the core tiny and lets it run the full 30–50-year life of the vessel on a single core.
• Ruggedization. The vessel, piping, and pumps must survive torpedo/mine shockwaves and work while pitching and rolling at speed, using advanced (classified) zirconium and nickel alloys against decades of saltwater.
• Rapid maneuvering. A sub evading a torpedo needs to throttle 10%→100% in seconds. Cores are tuned for a strong negative temperature coefficient so power rises naturally as steam demand cools the water.
• Acoustic silence. Noise equals death. Cores maximize natural circulation — hot water rising, cool water sinking — so noisy coolant pumps stay off except during high-speed sprints.

2.2Power & output vs. a commercial plant

Naval reactors are deeply classified, so public figures are authoritative estimates — but the picture of their density is vivid.

The Gerald R. Ford carries two A1B reactors — combined ~250 MW electricity (vs. ~64 MW on the older Nimitz class) plus 336,000 hp. That electrical surplus exists to run EMALS catapults, dual-band radar, and future directed-energy weapons. Total usable output approaches 500 MW — half an AP1000, packed into a fraction of the space. A reactor naming like S9G decodes as Submarine · 9th-generation core · General Electric (the designing lab).

2.3The HEU advantage

The magic of >93% HEU isn't a "more efficient" reaction — a single U-235 atom releases the same ~200 MeV anywhere. It's about power density and longevity.

Power density — more power per volume

Commercial fuel is ~95% U-238 "dead weight." A naval core is almost pure fissile material, so the core can be tiny → the pressure vessel can be small → the radiation shielding can be light. That cascade is the only reason a ~200 MWth reactor fits inside a 34-foot hull. Naval cores hit up to 200 MW/m³ vs. ~100 MW/m³ for a land PWR — double the density.

Longevity — overpowering the poison

Why doesn't commercial power use HEU? Two reasons: non-proliferation (93% material is weapons-grade — every plant would need military-base security) and economics (HEU is exponentially costlier; utilities want the cheapest kWh and have all the space they need for a bigger building).

2.4Who actually builds them

Not a Pentagon basement, and not an open commercial market. It's a textbook GOCO system — Government-Owned, Contractor-Operated — under Naval Reactors (the Naval Nuclear Propulsion Program), which holds dual citizenship in both the Navy and the Department of Energy. Its director is a 4-star admiral who is also a DOE deputy administrator, serving an unusual fixed 8-year term to guarantee zero political interference in nuclear safety.

• The labs. Bettis (historically Westinghouse) designed the first sub reactors; Knolls (historically GE) runs testing and the Kesselring land-prototype site where sailors train on real reactors before going to sea.
• The IP trap. Because the government funds 100% of R&D, it owns the designs. Contractors own the know-how — the tooling, metallurgy, and certified workforce. That expertise is exactly what's now feeding the civilian SMR boom.

3.1The civilian chain (LEU & HALEU)

The U.S. mines almost no uranium domestically. ~75–80% of Western yellowcake (U₃O₈) comes from Kazakhstan (Kazatomprom), Canada (Cameco), and Australia (BHP). Utilities buy ore on long-term contracts, pay an enricher for "Separative Work Units" (SWU), send the gas to a fabricator (Westinghouse, Framatome) for fuel rods, and own the product. It's entirely transactional.

3.2The military chain (>93% HEU)

Separated from the civilian market by a wall of national-security law. Military fuel must be "unobligated" — 100% American-mined, American-enriched, domestic tech only. The shocking part: the U.S. has no active full-scale facility making national-security HEU. The last federal enrichment plants closed decades ago.

No one buys or sells naval fuel — it remains sovereign U.S. property from vault to submarine to final repository. To rebuild domestic capability before the stockpile runs out, NNSA funds the DUECE program (Domestic Uranium Enrichment Centrifuge Experiment) with BWXT, targeting an all-American military centrifuge plant by the 2040s.

3.3The finite Cold War stockpile

The U.S. stopped enriching weapons HEU in 1964 and dedicated naval HEU in 1992. For 30+ years it has produced zero new HEU. Every active warhead and nearly every drop of naval fuel runs on atoms separated during the Cold War.

Critically, bomb cores are already ~93–94% — the same level the Navy needs. So Y-12 doesn't "downgrade" them; it recycles. Retired warheads (≈1,500 in the Pantex dismantlement queue) are stripped, the uranium is chemically purified to remove decades of decay impurities, melted, and re-cast as fresh feedstock. ("Down-blending" is a different program that dilutes HEU into 3–5% LEU or 5–20% HALEU for civilian and medical use.)

3.4The Kazakhstan tightrope

Kazakhstan is the Saudi Arabia of uranium — ~40% of world supply — and is historically tied to Russia, with virtually all exports once railed through Russia to St. Petersburg. After the Ukraine war, its ore stays available to the West because Astana engineered an escape hatch.

• The legal loophole. The Prohibiting Russian Uranium Imports Act bans Russian enriched uranium — not raw ore. Kazakh yellowcake enriched in Western plants (e.g. Urenco NM) is fully exempt.
• The corporate shield. Kazatomprom runs its mines as joint ventures with Cameco, Orano, China, and Japan — so Russia can't choke supply without a multi-superpower crisis (China especially depends on Kazakh ore).

4.1The two-stage architecture

The primary's only job is to create enough heat and X-rays to ignite the secondary, which produces most of the yield.

4.2The material recipe

Primary stage (the trigger)

• Weapon-grade plutonium (the pit). A hollow Pu-239 sphere, often with a thin HEU shell as a neutron reflector to shrink and lighten the core.
• Tritium + deuterium gas ("boosting"). Injected into the hollow pit just before detonation. Fusion of these gases floods the pit with neutrons, fissioning the plutonium far more completely — multiplying yield without a bigger bomb. The amount of gas injected is the knob behind "dial-a-yield."

Secondary stage (the main yield)

• Lithium-6 deuteride. The actual fusion fuel — a stable, inert, chalk-like solid. Designers don't store live tritium here; they breed it in-place during detonation.
• HEU spark plug + uranium tamper. A rod of HEU runs down the center; a heavy uranium cylinder wraps the outside to confine the fuel long enough to fuse — and then fissions itself for much of the yield.

4.3Radioactive vs. critical — a key distinction

A common misconception is that the parts are inert before assembly. They're intensely radioactive — yet physically incapable of a nuclear explosion on their own.

A fully assembled warhead is still radioactive — casings shield low-energy radiation, but gamma rays and neutrons leak out, so technicians use "stay-time" dose monitoring. Note the fusion fuel itself (Li-6 deuteride) is genuinely non-radioactive; the secondary is "hot" only because of the uranium spark plug and tamper sandwiching it.

4.4Tritium & limited-life components

Tritium has a 12.3-year half-life, decaying ~5.5%/year, so it's a Limited Life Component that must be periodically replaced or the weapon underperforms.

Routine gas swaps happen inside the active silo — a team retracts the 110-ton blast door, swaps the reservoir on the reentry vehicle, and the missile is back on alert within hours. A warhead only fully leaves the silo for major overhauls (via an armored Payload Transporter to the base Weapons Storage Area, and for total rebuilds, all the way to Pantex).

5.1Design & production sites

Design agencies (physics & engineering)

Production agencies (manufacturing)

5.2Pantex — where all roads lead

Pantex (Amarillo, TX) is the only U.S. site that assembles a weapon into finished hardware. Until it leaves Pantex, a warhead belongs to the DOE, not the military. Final assembly happens in Gravel Gerties — domed concrete rooms buried under hundreds of tons of gravel; if conventional explosives accidentally detonate, the roof blows up and the gravel collapses inward, trapping plutonium dust underground.

5.3Security & secrecy

Most defense secrets ride on Special Access Programs (SAPs). The nuclear complex goes further, under the Atomic Energy Act:

• "Q" Clearance. The DOE equivalent of DoD Top Secret, tailored to Restricted Data.
• Restricted Data (RD). Weapon design info is classified by law from the moment of creation — "born classified" — not by executive order.
• Sigma categories. Even with Q clearance, access to specific data is gated by Sigma number (e.g. Sigma 14 governs use-control/vulnerability data). You can hold the clearance and still be barred from a blueprint.

5.4Spotlight — BWXT

BWXT holds arguably the most tightly guarded secrets on Earth, yet is a fraction of the size of the aerospace primes (~$3.2B revenue vs. Lockheed's $65B+). Stealing a stealth-jet blueprint is a costly intelligence failure; stealing enrichment profiles or warhead triggers is a threat to global survival — hence security on an entirely different plane, and an unbreachable economic moat (zero traditional competition in its core sectors).

6.1The end of New START

New START — the last bilateral U.S.–Russia arms agreement — expired February 5, 2026, hitting its legal limit with no successor. For the first time in decades there are no binding caps on strategic arsenals.

How the 1,550 was divided — and why MIRVs matter

Because it was one shared cap, the Pentagon made trade-offs. The mobile, near-untrackable sea leg was prioritized; the fixed, vulnerable land leg was de-MIRVed to one warhead per missile to save "warhead slots."

The "bomber loophole": a B-52 carrying 20 cruise missiles counted as one warhead, since bombers aren't on hair-trigger alert and take hours to reach a target.

6.2Land — the Sentinel ICBM

The LGM-35A Sentinel (Northrop Grumman) replaces the 1970s Minuteman III — the biggest U.S. nuclear overhaul in history.

The plan

• 450 brand-new silos across North Dakota, Montana, Wyoming, Nebraska, Colorado — dug next to the old ones so Minuteman III stays on alert during the swap (refurbishing 55-year-old crumbling silos proved harder than building new).
• 400 active missiles in 450 silos — a deliberate "shell game": 50 silos sit empty and missiles shuffle between identical-looking holes, forcing an adversary to target all 450.
• Arms with the new W87-1 warhead in the Mk21A reentry vehicle, with backwards-compatibility to recycle existing W87-0 warheads as a production hedge.
• Net change vs. today: zero — a 1:1 technological swap (down from 1,000 Minuteman silos in 1967, cut by treaties to today's 450).

Timeline

It's now one of the largest infrastructure projects in modern U.S. history — 3,100 miles of new fiber, a prototype silo at Promontory, UT, and a construction nerve-center at F.E. Warren AFB.

6.3Sea — Columbia-class & the W93

The most survivable leg. Columbia-class SSBNs (GD Electric Boat + HII) replace the 1980s Ohio class. Lead boat USS District of Columbia is ~65% complete; delivery ~2028, first patrol ~2030, 12-boat fleet by ~2031.

Why a "smaller" submarine? (24 → 16 tubes)

Reinforcing reasons: a 24-tube boat would be too long for existing dry docks at Kings Bay and Kitsap, and the design was locked in during the optimistic post-2010 era when treaty caps made extra tubes pointless. (Now some in Congress push to expand from 12 to 16 hulls.)

The W93 warhead

• The 93rd U.S. design — a clean-sheet layout (built from previously tested nuclear components to avoid live testing), in a new Mk7 reentry body.
• Why it survived budget fights: the UK co-develops it (their "Astraea" shares the Mk7), and the Columbia "tube deficit" demands lightweight warheads — packing heavy W88s on a missile costs 1,300+ nautical miles of range.
• Post-treaty, funding surged past $800M; LANL ordered to double pit production to 60/year. Expected loading: 5–6 W93 per missile (≈96 per boat — matching an Ohio's treaty payload with four fewer tubes, while flying ~1,200 nm farther).

6.4Air — B-21 Raider & the LRSO

The flexible leg — bombers can be launched as a signal and recalled. The B-21 Raider (Northrop, Plant 42 Palmdale) replaces the B-1 and B-2. It completed an ultra-fast 73-day test phase; first jets arrive at Ellsworth AFB in 2027; planned fleet of 100–145 aircraft.

Two kinds of payload

• Propelled: AGM-181 LRSO. A new stealthy, Williams-turbofan cruise missile (Raytheon), 1,500+ miles, terrain-following — lets even a B-52 launch from safe ocean standoff. Carries the modernized W80-4 warhead.
• Unpropelled: B61-12 / B61-13 gravity bombs. No engine — the bomber must penetrate to the target, which is the entire reason for stealth. Modern B61s have guided tail-kits for near-pinpoint accuracy (B61-13 is a bunker-buster).

6.5The three new warheads compared

All three are two-stage thermonuclear weapons on identical Teller-Ulam physics — but tailored to entirely different doctrines.

Dial-a-yield on the W80-4 works by varying the volume of tritium/deuterium boost gas: minimal gas → incomplete burn → ~5 kt; maximum gas → full primary → 150 kt. For scale, Hiroshima was ~15 kt — a W87-1 at max is 30× that.

6.6The real bottleneck — plutonium pits

The U.S. is legally unconstrained but can't surge overnight. The binding limit is the 80 pits/year mandate — and the infrastructure is lagging.

Because SRS is late, 80/year by the early 2030s is out of reach — driving the Sentinel timeline slip and the W87-0 fallback hedge. Meanwhile the W80-4 cleverly sidesteps the pit problem (it reuses pits) but still strains Y-12 (fresh HEU secondaries and lithium) and jams Pantex, which is simultaneously running B61-12/13 production, W87-1 tooling, and W80-4 refurbishment.