The nuclear revival has crossed an important line: it is no longer being driven primarily by governments defending an unpopular technology. It is being pulled forward by customers who need enormous quantities of firm, carbon-free electricity, and who are finally willing to sign contracts that make new investment financeable.

That is the significance of the latest outlook from IAEA Director General Rafael Grossi, who said nuclear power could more than triple by 2060. The forecast itself is not the story. Nuclear forecasts have been revised upward before, and many have died on contact with construction schedules, financing costs, and public procurement. The important change is the demand behind this one. Electrification, artificial intelligence data centers, and energy security are creating buyers with a much lower tolerance for intermittency than the old renewable-only model assumed. More than 60 reactors are now under construction globally, while more than 400 operating reactors already produce about 10 percent of world electricity and roughly a quarter of its low-carbon electricity.

But the industry should not confuse a stronger demand signal with delivered capacity. A forecast is not a reactor, and a memorandum of understanding is not a project finance package. The same week that developers were discussing eight CANDU units in Indiana, six AP1000s in Kentucky, and new SMR deployments, Westinghouse reportedly cancelled commercial work on eVinci. That contrast is the nuclear market in one frame: serious customers are returning, but weak designs and unfunded business models are still being cleared out.

The revival is real. The easy version of the revival, in which every advanced reactor announcement becomes a construction program, is fantasy. The winners will be the companies that convert demand into permits, contracts, supply chains, and repeatable construction. Everyone else is selling optionality.

The most revealing nuclear deal in the current news cycle is not the IAEA projection. It is the agreement between Amazon and Constellation to support a $3 billion upgrade and relicensing effort at Maryland’s Calvert Cliffs plant.

This is what nuclear investment looks like when the commercial problem is understood correctly. Calvert Cliffs is not a speculative SMR. It is an operating plant with an established workforce, grid connection, fuel cycle, operating history, and two large reactors already producing electricity. Amazon’s 20-year power purchase agreement gives Constellation the revenue certainty needed to pursue another 20 years of operation and fund improvements. The project is expected to add approximately 190 MW of emissions-free capacity between 2030 and 2032, while the plant continues serving the regional grid rather than physically routing every electron to Amazon.

That last point matters. Corporate nuclear procurement is not primarily about building private wires from reactors to data centers. It is about underwriting dependable generation in a constrained regional power market. The buyer receives a credible claim on clean electricity and price certainty. The utility receives a long-duration customer. The grid receives more firm capacity. The arrangement is closer to infrastructure finance than to a conventional retail power purchase agreement.

The economics favor this kind of investment because existing nuclear assets have already absorbed the hardest capital costs. The marginal cost of extending a safe, productive reactor’s life is usually far below the cost of replacing its output with new firm generation. A relicensing and uprating program does not eliminate regulatory or outage risk, but it avoids the schedule risk that has damaged nearly every first-of-a-kind nuclear construction program in the West.

Amazon’s deal also exposes a weakness in the current SMR narrative. Data center demand does not automatically require an SMR. It requires electricity, quickly, at predictable prices, with a credible path through interconnection and permitting. An existing reactor upgrade can meet those requirements sooner than a reactor that has never been built. That does not make SMRs irrelevant. It means SMRs must compete not against an abstract fossil alternative, but against uprated large reactors, life extensions, grid-scale storage paired with generation, and new gas plants that can be built quickly.

The cost of nuclear remains dominated by financing and construction execution, not by uranium. A reactor that operates at a high capacity factor for 60 or 80 years can produce enormous quantities of electricity, but that advantage disappears when construction takes too long or the cost of capital compounds during delays. The commercial value of the Calvert Cliffs transaction is therefore not just the energy it buys. It is the risk it removes from the capital stack.

That is the template the IAEA forecast needs. Tripling nuclear capacity requires more than favorable public opinion and more than ambitious national targets. It requires contracts that make lenders believe the revenue will still exist after the ribbon cutting, and industrial systems that can build multiple units without relearning the same lessons each time.

First, it accelerates the life-extension economy. Constellation has already secured nuclear agreements with major technology and retail companies, and the Calvert Cliffs transaction gives the market a concrete example of what those agreements can finance. The immediate beneficiaries are operators with good plants in high-demand markets, especially where transmission constraints make local firm generation unusually valuable.

Second, it strengthens the case for standardized large reactors. The proposed Indiana campus of up to eight 700 MW CANDU reactors is still an exploration agreement, not an approved build, but its scale is strategically important. A fleet approach offers the possibility of spreading engineering, procurement, training, and regulatory costs across multiple units instead of treating every reactor as a bespoke monument. The same logic applies to the proposed six-unit AP1000 effort in Kentucky. The headline capacity is less important than whether the developers can turn a multi-unit ambition into a repeatable construction program.

Third, it forces SMR developers to separate technical progress from fundraising. Antares’ approval of a Nuclear Safety Design Agreement for its Mark-1 reactor is a meaningful step because it establishes design requirements, safety analysis methods, and regulatory expectations for the project’s next phase. But it is still a design and authorization milestone, not operating experience or commercial validation. Investors should ask what follows: a site, a customer, a detailed cost estimate, a supply-chain contract, and a construction schedule with contingency. If those pieces are absent, the milestone is valuable engineering work, but not yet an energy business.

The countries moving fastest will be those that combine demand certainty with institutional competence. The United States has the customer base and a deep industrial ecosystem, but it remains vulnerable to permitting delays, fragmented regulation, and first-of-a-kind cost overruns. Poland is building political and commercial momentum around BWRX-300 deployment, including a proposed six-unit project at Pątnów, but it still must demonstrate that the fleet can move from joint venture structure to financing and construction. Bangladesh’s Rooppur Unit 1 is nearing power startup, subject to final regulatory clearance, a reminder that a nuclear newcomer can reach the finish line, but also that commissioning is governed by evidence, not ceremony.

The losers will be companies whose strategy depends on the market rewarding novelty before it rewards delivery. Westinghouse’s eVinci cancellation is a warning that a technically attractive microreactor concept can still fail to become a commercial product. The nuclear industry does not need more PowerPoint reactors. It needs designs that can survive a customer’s procurement process, a regulator’s review, a lender’s diligence, and a construction site.

The IAEA is right that nuclear could more than triple by 2060, but the path will not look like a single technology boom. It will look like a layered buildout: life extensions first, uprates where the grid can absorb them, standardized large reactors in countries with serious state capacity, and SMRs only where their smaller scale solves a real siting, financing, or industrial problem.

The nuclear revival is therefore real progress, not noise. But its center of gravity has shifted. The question is no longer whether the world wants more nuclear electricity. Data centers, manufacturers, utilities, and governments are answering that question with money. The question is which developers can turn that demand into bankable megawatts before the next customer chooses gas, imports electricity, or moves its load elsewhere.