Tech companies are bringing shuttered nuclear reactors back online specifically to power AI data centers, a strategy that sidesteps the multi-year grid connection bottleneck strangling energy infrastructure. Microsoft is spending roughly $1.6 billion to restart Three Mile Island Unit 1 in Pennsylvania, while Holtec International is attempting the first-ever full restart of a reactor already in decommissioning at Palisades, Michigan. These aren't one-off deals; they represent a fundamental shift in how hyperscalers solve their energy crisis.

Why Are AI Data Centers Consuming So Much Power?

The numbers explain everything. A traditional server rack draws around 5 to 10 kilowatts, roughly equivalent to powering a few large homes. An AI-optimized rack packed with graphics processing units (GPUs) pulls 50 to 100 kilowatts, about ten times more power in the same physical footprint. A single large training cluster running 50,000 or more high-end GPUs can draw 150 to 200 megawatts continuously, which is close to what a mid-sized city uses.

The scale is accelerating rapidly. According to the International Energy Agency's Electricity 2026 report, global data center electricity use is on track to roughly double from about 485 terawatt-hours in 2025 to near 950 terawatt-hours by 2030. In the United States alone, data centers could consume between 9 and 17 percent of national electricity by the end of the decade, up from around 4 to 5 percent today.

What's Blocking New Power Plants from Being Built?

The obvious solution would be to simply plug into the existing grid and build more solar or natural gas plants. In practice, that queue is completely jammed. More than 2,500 gigawatts of generation and storage projects are currently stuck waiting for grid connection worldwide, according to the International Energy Agency. Building new transmission infrastructure typically takes 5 to 15 years, far longer than constructing the generation itself. In parts of the United States, interconnection wait times now average close to five years, with some regions quoting seven to ten years.

This mismatch between how fast AI companies want to build and how slowly grids can expand is the core problem. It's a physical-world bottleneck that no amount of software engineering can solve. Tech companies need power now, not in a decade.

Why Decommissioned Nuclear Plants Are the Shortcut

Shuttered reactor sites offer something almost impossible to build quickly today: existing high-voltage transmission lines already running into the regional grid, industrial-grade cooling water access, land already zoned for heavy generation, and in many cases a license that can be reinstated rather than applied for from scratch. These advantages compress timelines dramatically.

The clearest example is Three Mile Island Unit 1 in Pennsylvania, shut down in 2019 for economic reasons and sitting next to the infamous Unit 2 that partially melted down in 1979. Constellation Energy is spending roughly $1.6 billion to bring it back online, rebranded as the Crane Clean Energy Center, under a 20-year power purchase agreement with Microsoft reportedly worth around $16 billion. The plant is expected to reach commercial operation around 2027, generating about 835 megawatts dedicated almost entirely to Microsoft's AI data centers.

The more dramatic example is Palisades, an 800-plus megawatt plant in Covert, Michigan. Holtec International originally bought Palisades in 2022 to decommission it, then reversed course once demand forecasts from AI and advanced manufacturing started climbing. It's now attempting something that has literally never been done in US history: restarting a plant after it had already entered decommissioning. The company has missed its own restart date twice already, which says a lot about how unpredictable this process really is.

How to Understand the Advantages of Restarting Old Nuclear Plants

• Transmission Already Exists: No multi-year grid interconnection queue. The high-voltage lines are already in place and connected to the regional grid, eliminating one of the biggest bottlenecks facing new power plants.
• Cooling Water Systems Permitted: Cooling water intake infrastructure is already permitted and operational. New plants must navigate years of environmental reviews; old sites skip this step entirely.
• Zoning and Community Acceptance: The site is already zoned and community-accepted for heavy industrial power generation, avoiding local opposition that often delays new projects.
• Reusable Infrastructure and Licenses: Some equipment, staff expertise, and even the original operating license can sometimes be reactivated instead of rebuilt from scratch, saving time and money.
• High Capacity Factor: Nuclear runs at a capacity factor above 90 percent, more than double what most solar or wind farms achieve on their own, meaning more reliable, consistent power delivery.

What Does It Actually Take to Restart a Dead Reactor?

Announcing a nuclear restart deal and actually flipping the switch are two very different timelines. Restarting a plant that has gone through decommissioning is not like flipping a breaker back on. The process involves multiple complex steps that can take years.

First comes engineering and safety assessment, inspecting the reactor vessel, steam generators, and containment for corrosion or degradation after years sitting idle. Then the US Nuclear Regulatory Commission (NRC) must reinstate the license; the agency actually built an entirely new inspection framework, called IMC 2562, just to oversee this kind of restart, since it had never happened before. Physical refurbishment follows, replacing reactor head penetrations, cleaning and passivating primary systems, and reinstalling turbines and generators that were partly stripped for parts or scrap.

New nuclear fuel must be delivered and inspected, and staff need to be re-certified by industry bodies like INPO (Institute of Nuclear Power Operations). Environmental and water permitting comes next; Holtec's own filings show Palisades alone is expected to draw around 25,000 gallons of water per minute once its planned small reactor units are added, which triggers its own separate review. Finally, grid interconnection and power purchase agreement finalization must be completed, even with existing transmission lines already in place.

Holtec has reportedly listed more than 5,000 individual remaining work items for Palisades even after major milestones were reached. That's the behind-the-scenes reality that press releases tend to skip. The process is routine but far from trivial, and delays are common.

How Does This Compare to Other Power Options?

Tech companies have multiple paths to secure power, each with different timelines and tradeoffs. Restarting a decommissioned reactor takes 2 to 4 years in the optimistic case, though delays are frequent. Once running, reliability is very high, but the biggest risk is unknown equipment condition and repeated delays, as Palisades has already experienced.

Building a new small modular reactor (SMR) takes 5 or more years, and commercial-scale deployment remains unproven. Reliability is expected to be high, but first-of-a-kind construction risk is significant. A new gas turbine plant can be built in 1 to 3 years with high reliability, but it generates emissions that conflict with corporate climate targets. Solar and wind with battery storage can be deployed quickly but require grid interconnection, which brings us back to the original bottleneck.

Against this backdrop, restarting old nuclear plants looks like the fastest path to reliable, carbon-free power for AI companies willing to navigate the regulatory complexity. It's not a perfect solution, but it's the one that actually works within the current grid constraints.

The broader implication is clear: AI's power hunger is reshaping energy infrastructure in real time. Companies are not waiting for policy to catch up or for new transmission to be built. They're buying their way into existing assets and bringing dead power plants back to life, one bolt at a time.