The coverage cliff, and the date it lands on

Every load-bearing date in this note is a political date. The coverage gap is a contracting fact, but the reason it is dangerous is that it opens in the same year a statutory authority extinguishes. What follows is the policy layer, actor by actor, with the instrument named in each case. Our conviction on political outcomes is graded lower than on the market analysis above, and the calls here should be read that way.

1. United States

   The cliff is the waiver, not the ban

   The Prohibiting Russian Uranium Imports Act bans imports to the end of 2040, but DOE's authority to waive that ban terminates on 1 January 2028 P. The ban is the durable instrument; the waiver is the one that expires, and it expires first. No successor mechanism has been disclosed, and DOE publishes no register of waivers, recipients, or quantities U, so the market cannot size what is about to stop even if it wanted to.

   The political economy runs both ways. Domestic enrichment build-out is the constituency that benefits from letting the authority lapse; utilities and their ratepayers are the constituency that benefits from an extension. Both are organised, and neither has had to fight yet because the date is still far enough away.

   Read. We expect the authority to be extended or replaced rather than allowed to lapse cleanly, because the alternative imposes a visible cost on regulated utilities in an election-adjacent year. But we hold that view at low conviction and we would not build a coverage model on it. The planning case is the statute as written: the authority ends on 1 January 2028, in the same year uncovered requirements reach 17.8%.

2. European Union

   The exposure sits in the link no sanctions package has touched

   Successive EU sanctions packages have repeatedly excluded nuclear fuel, and there is no EU-level prohibition on Russian enrichment or conversion services. The result is visible in the Euratom data: EU enrichment was 22.55% Russian and conversion 24.40% Rosatom in 2025 P. Conversion is where the EU is genuinely more exposed than the United States, and it is precisely the link that no political instrument currently addresses.

   The blocking constraint is that several member states operate VVER reactors with fuel-cycle arrangements that cannot be substituted on a sanctions timetable. Unanimity plus a physical dependency is a reliable recipe for carve-outs.

   Read. We do not expect an EU prohibition on Russian conversion or enrichment within the forecast horizon. The EU's 42,522 tU of inventory, described as more than three reloads P, is what a bloc holds when it has decided to buy time rather than force the issue.

3. Russia

   The leverage can be exercised from the supply side first

   Rosatom holds 42.8% of world enrichment capacity as a single operator P, and supplied more enrichment to US utilities in 2025 than domestic capacity did, at 25.8% against 23.1% P. Every Western policy discussion treats the decision as the buyer's. It is not obviously the buyer's.

   A supplier with two years of visibility on the expiry of its customers' legal cover has the option to withdraw on its own schedule, at a moment when the replacement capacity is not yet qualified. Export restrictions have already been used as an instrument elsewhere in this research series.

   Read. The asymmetric risk in 2027 is not that Washington fails to extend the waiver. It is a Russian restriction landing before the waiver question is settled, which would compress a two-year adjustment into a procurement cycle. We rate this unlikely but materially under-priced.

4. Niger and France

   Resource nationalism arrived and the market did not move

   SOMAIR output fell from 2,020 tU in 2022 to 960 tU in 2024, Orano lost operational control in December 2024, Niger nationalised the operation in June 2025, and Orano filed arbitration in January 2025 P. Niger supplied 33 tU, or 0.23% of EU deliveries in 2025 P, against a historic position as a core French supplier.

   Read. The useful lesson is about sequencing rather than about Niger. The dependency was unwound on the demand side before the political dispute resolved, which is why an expropriation in a formerly strategic supplier produced no price event. Apply the same test to Kazakhstan: the question is not whether political risk exists, it is whether buyers have already routed around it. On Kazakh volumes, they plainly have not.

1. Does the coverage gap actually have to be filled?

   The sceptical view is that maximum anticipated requirements are a ceiling rather than a forecast, that utilities routinely run thin coverage a decade out and fill it opportunistically, and that the 2034 number will look nothing like 94% uncovered by the time 2034 arrives.

   Our answer. Correct in principle and less comforting than it sounds. Utilities have always carried an out-year gap; what is different is that the near years are tightening at the same time. Coverage at five years out, in 2031, is already only 30.6%, and the seller-side book at the largest Western producer thins on the same schedule. The gap does not have to be filled on EIA's timetable. It does have to be filled before the fuel is loaded.

2. Do reactor restarts and datacenter power deals move fuel demand?

   Palisades began fuel loading on 31 August 2026, Crane targets 2027 with a twenty-year corporate PPA behind it, and Duane Arnold targets 2029 with a twenty-five year agreement. The narrative writes itself.

   Our answer. Materially less than the narrative implies, and we would tell a client so. Those three units total 2,234 MWe. At the roughly 196 tU per GWe-year implied by the US fleet ratio, that is about 438 tU a year, or 1.14 million pounds of U₃O₈ once all three run, plus a first core of two to three annual reloads KR. Against 46.9 million pounds of annual US purchases that is around 2.4%. It is real and it is not transformative. Note also that the power purchase agreements commit megawatt-hours, not fuel: neither discloses any fuel supply commitment.

3. Is Kazakh supply the swing, and what is actually constraining it?

   Kazatomprom guided 2026 production to 27,500 to 29,000 tU on a 100% basis and cut revenue guidance at its first-half results on 21 August 2026. The market reads this as price-driven discipline.

   Our answer. Part of it is an input constraint, and it connects to another of our editions. In-situ recovery consumes sulphuric acid, and the company disclosed that commissioning of its TQZ sulphuric acid plant has slipped from the first quarter of 2027 to a window of Q3 2027 to Q1 2028 P. Sulphur is simultaneously the binding constraint on world phosphate production. Two commodities that look unrelated are competing for the same molecule, and the uranium market is not pricing it because uranium analysts do not read fertiliser data.

4. Is enrichment concentration the right thing to worry about?

   The Russian import ban addresses roughly a quarter of US SWU purchases, and $2.7 billion of DOE money plus the Urenco and Orano expansions are the answer.

   Our answer. The trade flows understate the structural problem. Rosatom holds 42.8% of world enrichment capacity PKR, against Urenco at 25.4% and Orano at 11.0%. The domestic build-out is real and it lands in 2029 or later on every announced schedule, which means the 23% US-origin share in this note will not move materially before then. The honest framing is that the policy problem was created in 2024 and the capacity answer arrives at the end of the decade.

• EIA's 2026 Uranium Marketing Annual Report, due July 2027, showing unfilled 2031 requirements below 15,000 thousand pounds. That would say the contracting cycle turned a year earlier than the current book implies and would take the urgency out of our lead.
• DOE publishing a waiver register, or announcing a transition mechanism for committed Russian deliveries beyond 1 January 2028. Either removes the correlation between the policy date and the contracting date.
• A current, operator-sourced conversion capacity and throughput table. Our conversion thesis rests on a 2022 dataset and we would revise it in either direction on better data.
• Cameco's contract book flattening rather than declining across 2029 and 2030 in a subsequent MD&A. That is the seller-side confirmation of our buyer-side series, and its removal would weaken the argument considerably.
• Kazatomprom guiding 2027 production above 30,000 tU on a 100% basis, or confirming TQZ commissioning back in the first half of 2027. Both would ease the acid constraint we have flagged.
• Sustained US-origin enrichment above 30% of purchases in a subsequent EIA release. That would mean the domestic build-out is arriving earlier than every announced schedule implies.