The nuclear revival is no longer a forecast problem. It is becoming a procurement problem, and that is real progress.

The clearest evidence is not Rafael Grossi’s prediction that global nuclear capacity could more than triple by 2060, although the drivers are credible: electrification, artificial intelligence data centers, and technology companies demanding firm, carbon-free power. It is the harder-to-fake activity underneath that forecast. Holtec, Constellation, and NextEra are working to return three closed American reactors to service, potentially restoring 2.25 GW, with Microsoft and Google supporting projects through long-term power agreements. The U.S. Department of Energy is prepared to provide up to $4.4 billion in loans. That is not a press release about a conceptual reactor. It is capital being assembled around existing machines, existing sites, existing grid connections, and known operating histories.

The distinction matters. Nuclear has spent the past decade producing an impressive volume of announcements and a much smaller volume of concrete megawatts. Restarting Palisades, the former Three Mile Island Unit 1, and Duane Arnold does not solve the industry's construction-cost problem. It does something more immediately valuable: it tests whether customers will pay enough for reliable nuclear electricity to justify expensive work on assets that already possess the attributes new projects spend years trying to obtain. Palisades is targeting a return around March 2027, while Constellation is investing roughly $1.6 billion to restart its Pennsylvania unit and Duane Arnold is targeting the first quarter of 2029.

The hype is wrong if it treats these restarts as proof that new nuclear is suddenly cheap. The fear is wrong if it treats them as a bailout of obsolete plants. These projects are a market test. If large technology buyers will sign contracts long enough and strong enough to support refurbishment, fuel replacement, licensing, and grid work, nuclear’s value is not theoretical. It is being priced into contracts. The uncomfortable question is whether that value can extend beyond a handful of unusually attractive sites.

The economic case begins with what these projects are not paying for. A restart avoids much of the cost and schedule risk associated with greenfield nuclear construction. The site already has transmission infrastructure, cooling systems, security arrangements, a trained regional workforce, an emergency planning framework, and a reactor whose design and operating record are known. Those advantages do not make refurbishment cheap, but they compress the risk stack.

Palisades is an 800 MW pressurized-water reactor that closed in 2022. Its restart is now in the refueling phase, though fuel-handling work has encountered delays. Constellation’s Crane Clean Energy Center, the renamed former Three Mile Island Unit 1, represents a different commercial model, with Microsoft’s 20-year power purchase agreement providing a visible customer anchor. Duane Arnold, a 615 MW boiling-water reactor, adds a third test case, though its 2029 target shows that even a restart is measured in years, not quarters.

The relevant comparison is not nuclear against the lowest quoted solar or wind energy price. It is nuclear against the cost of delivering electricity when the grid actually needs it. A reactor operating at a high capacity factor supplies energy through nights, weather events, and periods when data centers cannot curtail load. Replacing that service with a portfolio of variable generation requires overbuilding, transmission, storage, dispatchable backup, or some combination of all four. Those costs are often excluded from simplistic levelized-cost comparisons.

That does not give nuclear a free pass. A reactor that is expensive to refurbish, poorly located, or unable to secure a bankable offtake contract can still be uneconomic. The Microsoft agreement is therefore more important than its symbolism. It transfers part of the reliability premium from a regulated utility rate base to a creditworthy corporate customer with an urgent load problem. Technology companies are not buying nuclear because they have become nuclear advocates. They are buying firm power because an intermittent supply profile is incompatible with expanding artificial-intelligence workloads.

The restart economics also expose a weakness in the new-reactor narrative. An SMR may eventually reduce construction risk through repetition and factory production, but no factory benefit exists before there is an orderbook. The first unit still carries engineering, licensing, supply-chain, financing, and schedule risk. An operating reactor returning to the grid bypasses much of that learning curve. For the next several years, the fastest route to more nuclear generation is likely to be preserving, restarting, and uprating existing plants while new designs fight to establish commercial evidence.

First, it accelerates the commercial case for nuclear power purchased directly by large loads. Microsoft’s agreement with Constellation gives other data-center operators a reference point. Google, Amazon, and industrial customers now have a credible alternative to building ever larger portfolios of renewables and backup generation around constrained transmission connections. The winners will be utilities that can offer firm nuclear output with clear schedules, credible operating records, and contracts that allocate outage and refurbishment risk honestly.

Second, it strengthens the position of operators with dormant or threatened assets. Holtec’s Palisades work is especially important because it links a restart to a future SMR strategy. The U.S. Department of Energy identifies Holtec’s plan to deploy two SMR-300 reactors at the Palisades site as part of its Generation III+ program. That creates a potential progression from restoring an existing plant to developing a repeatable new-build platform, but the order matters. A successful restart builds institutional confidence and site value. It does not automatically prove that the SMR will be competitive.

Third, it improves the prospects for countries that treat nuclear as an industrial capability rather than a single megaproject. The IAEA says more than 400 reactors in over 30 countries produced roughly 10% of global electricity in 2025, while more than 60 reactors were under construction. Its forecast of more than tripling nuclear use by 2060 reflects a major change in demand assumptions, but forecasts do not create supply chains. The countries positioned to capture the growth are those securing heavy manufacturing, fuel services, qualified labor, regulatory capacity, and repeat orders now.

The companies most at risk are not necessarily those with the least exciting reactor designs. They are the vendors and developers that confuse regulatory progress with commercial progress. A construction permit, including the NRC’s approval for TVA’s 300 MW Clinch River SMR, is a meaningful milestone, but it is not an operating plant, an electricity contract, or proof of cost. TVA still needs a separate operating license before the reactor can be fueled and generate power. Nuclear executives should stop presenting that distinction as a technicality. It is the difference between permission to build and evidence that the business works.

The next signal to watch is not another capacity projection. It is whether the three restart projects hit fuel-loading, grid-connection, and commercial-operation milestones without turning their loan support into open-ended subsidy. If Palisades returns on schedule, Crane secures its long-term economics, and Duane Arnold progresses toward 2029, the market will have something more valuable than optimism: a repeatable playbook for converting retired nuclear assets into firm clean-power supply.

That is where the industry is heading. The nuclear revival will not be won by the company with the most ambitious reactor rendering. It will be won by the companies and utilities that turn reliability into a contract, a contract into financing, and financing into megawatts. The first wave is not glamorous. It is refurbishment, restarts, uprates, and disciplined procurement. But that is how a revival becomes an industry.