CorrectionThe 17 July Daily Briefing described a ~20% fall in European gas that did not happen — August TTF settled at €54.79/MWh on 16 July, essentially flat. During our platform rebuild, a retired machine running an outdated data feed briefly came back online and republished week-old settlements as live prices. The briefing has been withdrawn, and live prices are now verified against exchange settlement history before publication.
Nuclear ETFs Re-Rate as Hyperscalers Lock In Long-Term Reactor Contracts
Microsoft's Three Mile Island restart deal has set a template that uranium equity funds are now pricing across the US and Asia.
The Global X Uranium ETF (URA), which holds $6.86 billion in assets, closed at $38.73 on Friday (2026-07-17), down 0.74% on the session but higher over the past year as direct power purchase agreements between hyperscalers and reactor operators have changed the demand structure for nuclear capacity.5 The clearest single transaction came in September 2024, when Microsoft signed a 20-year, 835-megawatt power purchase agreement with Constellation Energy to restart Three Mile Island Unit 1, a $1.6 billion project targeting first generation in 2027.5
That deal is not an isolated event. US power generation from data centres is projected to climb from around 5% of the national total to roughly 15% over a five-year span, according to analysis published in late May 2026 — a step change on a grid that has barely grown in capacity since 2000.5 Nuclear can absorb that load in ways intermittent generation cannot: a 1-gigawatt reactor runs at capacity factors above 90% and occupies a fraction of the land required by an equivalent solar build.5
Japan is facing the same pressure from a different direction. Wood Mackenzie estimated in May 2026 that Japanese data centres will consume electricity equivalent to 15 million to 18 million households by 2034, driving 60% of the country's total power demand growth as hyperscalers commit capital at scale.3 Oracle, Google, and Microsoft were designated official cloud partners by the Japanese government, with combined investment of $28 billion (4 trillion yen).3
That investment profile creates a secondary nuclear bid in Japan. The government's cloud-partner designations signal a commitment to underwriting base-load supply on a decade-long horizon, and the restarting domestic reactor fleet is the most credible source of that supply.3 Inpex, Japan's largest oil and gas producer, operates in an energy market where power policy is shifting toward always-on generation rather than peak capacity alone.5
The three main funds offer distinct access points. NUKZ returned approximately 52% over the past year through late May 2026 and roughly 11% year to date, with shares near $71.5 URAN, with roughly $841 million in assets and an expense ratio of 0.85%, provides direct exposure to Japanese and Korean reactor builders at lower cost than some alternatives.5 URA offers the deepest liquidity alongside pure uranium-price exposure.5 Each fund is a different entry into the same thesis: that AI infrastructure is a structural demand event for nuclear, not a cyclical trade.
South Korea is developing the same logic at a military level. In late May 2026, Defense Minister Ahn Gyu-back unveiled a roadmap for nuclear-powered submarines, adding state demand for reactor expertise to the civilian buildout already under way.6 The scale of that programme is smaller than data centre demand, but its emergence signals how broadly Seoul is treating nuclear competence as a strategic industrial capability.
On the supply side, the build timeline may compress. Idaho National Laboratory partnered with NVIDIA on a project called Prometheus, aiming to cut reactor development times by up to 50% and reduce operating costs by a similar margin through AI-assisted design.4 If the targets are met, that schedule shift would directly expand the addressable market for the reactor manufacturers held by URAN.4
The US government's policy direction is unambiguous. Washington is targeting a quadrupling of nuclear capacity from roughly 100 gigawatts in 2024 to 400 gigawatts by 2050, a 25-year programme that will require sustained permitting and capital at a scale the industry has not managed since the 1970s.2
Capital is moving ahead of that timeline. In May 2026, investment was rotating toward energy companies capable of supplying power for AI data centre buildouts, with nuclear and renewable baseload seen as the most reliable solutions to supply constraints that repriced Fluence Energy's shares 98.2% in a single week after the company disclosed master supply agreements with two hyperscalers and a record $5.6 billion backlog.1 For uranium fund holders, the immediate test is execution: whether reactor restarts and new builds arrive quickly enough to honour contracts hyperscalers have already signed, and whether uranium supply can scale to meet a demand cycle now anchored by industrial customers carrying 20-year commitments.1