Google's Nuclear Bet Rewires Big Tech's Energy Playbook
Google's 20-year nuclear power deal with Constellation Energy is more than a supply contract—it's a signal that big tech has exited the era of waiting for grid upgrades and is instead funding the build-out itself. What happens next reshapes utilities, uranium markets, and the politics of AI energy.
The Deal That Makes Sense Only Today
Google and Constellation Energy announced a 20-year power purchase agreement that, on its face, looks like another corporate green energy contract. But the structure is what matters. Constellation will invest $4.3 billion to modernize equipment at 11 existing nuclear plants across Illinois, Pennsylvania, and New Jersey, boosting capacity by 890 megawatts—enough to power the equivalent of a new reactor—without actually building one from scratch. The upgrades target turbines, steam generators, and digital control systems, delivering output comparable to several small modular reactors but on a compressed timeline. Capacity gains should start flowing by 2028.
Constellation will also supply Google with an additional 2,700 megawatts under a separate 15-year agreement. Crucially, the 890 megawatts of unlocked capacity feeds the PJM grid generally rather than being wired directly to a single data center. That distinction is the entire story.
Why Old Reactors Just Became S&P 500 Heroes
Constellation stock surged nearly 12% at Tuesday’s open, opening above its 200-day moving average for the first time since January. It was the fastest-moving stock in the S&P 500 that morning. Alphabet dipped 0.3%, still trapped in a 20-week consolidation pattern. The market is pricing in something Google investors are not yet fully registering: the nuclear renaissance is no longer speculative. It is a revenue contract.
This deal arrives barely a month after Amazon closed a similar 20-year agreement with Constellation, securing 690 megawatts from the Calvert Cliffs plant in Maryland, including 190 megawatts of new capacity. Amazon’s deal alone unlocked over $3 billion in infrastructure investment and, perhaps more importantly, provided the revenue certainty Constellation needed to relicense that plant for another two decades. Both deals together represent roughly 4,580 megawatts of committed nuclear-sourced power from a single provider—a scale of corporate procurement the industry has never seen.
The implications for Constellation’s balance sheet are immediate. The company now carries two long-duration, investment-grade counterparty contracts that anchor its forward revenue visibility through the early 2040s. That visibility is precisely what licensing extensions and capital projects require. Nuclear regulators do not hand out 20-year renewals based on hope; they respond to demonstrated cash flow. Google and Amazon have effectively underwritten Constellation’s near-term strategy.
The PJM Problem No One Is Solving Quietly
The reason this matters beyond stock tickers is the grid itself. PJM Interconnection, the regional transmission organization that serves 13 states and the District of Columbia, is simultaneously the busiest electricity market in the US and the one most exposed to the collision between surging data center demand and aging infrastructure. Last Wednesday, Constellation and NRG Energy stocks fell after the Federal Energy Regulatory Commission imposed a five-month hold on PJM’s plan to procure 6.8 gigawatts of new capacity—a move FERC characterized as shifting costs to rate-paying consumers.
Rate increases for PJM households have already sparked political pushback. Pennsylvania Governor Josh Shapiro moved recently to give local communities veto power over new data center projects, a direct response to constituent anger over electricity bills. The Google-Constellation deal is a work-around to that bottleneck: rather than waiting for regulatory approval of new grid capacity or facing community opposition to new construction, Google is funding efficiency upgrades at existing plants and buying the output long-term.
PJM’s capacity auction mechanics were designed for an era when demand grew slowly and predictably. Today, data centers in northern Virginia alone are projected to draw more electricity than the entire state of Maryland by 2028. The grid operator has struggled to reconcile that trajectory with transmission constraints that can take a decade to resolve. Corporate power purchase agreements like Google’s represent a parallel market emerging outside PJM’s traditional procurement framework—one that prioritizes speed and certainty over the regulated model’s cost-sharing principles.
The Second-Order Effects Will Compound
The deal’s ripple effects extend well beyond Constellation’s shareholders. Uranium producers are already repricing on renewed nuclear optimism, but the immediate beneficiary remains the plant operator, not the fuel supplier—because upgrading an existing reactor does not proportionally increase uranium consumption. The marginal cost of additional megawatts from a modernized unit is largely fixed, which means the profit margin on Google’s 2,700-megawatt purchase is wider than the headline numbers suggest.
Traditional utility stocks that have priced themselves as the default beneficiaries of data center demand may find their thesis undercut. If hyperscalers can secure direct corporate power purchase agreements at existing nuclear assets, the argument for waiting on utility-led transmission buildouts weakens. Regional distribution cooperatives in PJM states could face a structural disadvantage: they serve rate-paying customers who absorbed the cost of grid expansion while tech companies locked in supply elsewhere.
The labor market dimension is also significant. Constellation’s investment at 11 plants creates roughly 7,200 construction jobs and sustains 4,400 existing ones, according to the company. Those numbers concentrate in states where nuclear employment has been declining for years. Pennsylvania and Illinois both have populations that voted heavily for trade protectionism and industrial policy—sectors that now find themselves aligned with a technology company’s energy strategy in ways neither side likely anticipated.
Who Wins, Who Loses, What Comes Next
Constellation wins immediately. The company now has two blue-chip tech customers locking in revenue for 15 to 20 years, which de-risks the licensing extensions and capital projects that define its near-term outlook. BMO Capital reiterated an outperform rating on CEG stock following the Amazon deal, noting that the longer duration and premium to forward pricing drove both near- and long-term value. The Google deal likely reinforces that thesis.
Google wins on timing. Building a new nuclear plant takes a decade or more even under the most favorable regulatory conditions. Upgrading existing reactors takes three to five. For a company whose AI ambitions are constrained by power availability, that gap is existential. The 2,700-megawatt supply agreement gives Google operational certainty without waiting for SMR deployment or new plant permits.
The losers are less obvious but real. Traditional utility stocks that have priced themselves as the default beneficiaries of data center demand may find their thesis undercut if hyperscalers bypass the grid entirely and strike direct corporate power purchase agreements at existing nuclear assets. Regional grids like PJM face a structural question: if the largest buyers of electricity are securing off-take agreements directly with generators, what happens to the procurement model that FERC just paused?
The more significant shift is ideological. Big tech has stopped treating nuclear power as a niche sustainability credential and started treating it as core infrastructure, the same way it treats silicon and bandwidth. The market is only now catching up. Google’s bet signals that the next phase of AI expansion will be determined not by chip density or model architecture but by who controls the megawatts—and the companies that move first will define the terms for everyone else.