Hyperscalers building AI data centers are locking in power supply through multi-decade agreements with nuclear utilities, turning historically defensive energy stocks into AI growth plays.
Data center power demand is projected to more than double from 540 kilowatt-hours per capita in 2024 to 1,200 kWh per capita by 2030, but grid infrastructure development lags behind data center construction by years, creating a critical supply bottleneck that favors established power producers like Constellation Energy and Vistra Energy.
What happened
Major cloud providers are locking in energy supply through multi-decade power purchase agreements with nuclear utilities. Constellation Energy signed roughly 920 MW of long-term nuclear contracts averaging 18.5 years in the second quarter, while Vistra signed a 2,600 MW agreement with Meta Platforms and up to 1,200 MW with Amazon Web Services. Vistra also partnered with KKR, Nvidia, and the Kuwait Investment Authority to form Helix Digital Infrastructure with $10 billion in capital commitments.
Why it matters
Energy infrastructure has become a bottleneck for AI expansion—while data centers take up to two years to build, grid infrastructure can take four to 10 years or longer. Data center power consumption is projected to rise from about 540 kilowatt-hours per capita in 2024 to 1,200 kWh per capita by 2030. Energy stocks traditionally viewed as defensive, low-growth investments are now positioned as AI growth plays because hyperscalers need reliable baseload power and are prioritizing carbon-free nuclear or low-carbon sources to meet decarbonization mandates.
What to watch
Constellation and Vistra, both down 32% from their 52-week highs, face regulatory scrutiny around co-location or behind-the-meter deals and carry the risk of AI capex drying up, which would reduce energy demand growth projections.
The explosive growth of artificial intelligence data centers is creating an unexpected investment narrative: energy companies, traditionally viewed as slow-growth defensive stocks, are becoming plays on AI capex. The driver is straightforward and structural—hyperscalers building data centers consume enormous amounts of electricity, but the time required to expand the power grid lags far behind construction timelines.
While a data center can be built in up to two years, developing the necessary grid infrastructure takes four to 10 years or longer. Demand is accelerating rapidly: according to the International Energy Agency, data center power consumption averaged about 540 kilowatt-hours per capita in 2024, with projections showing it could rise to 1,200 kWh per capita by 2030. This structural supply-demand mismatch has made power capacity itself a scarce and valuable resource.
Constellation Energy and Vistra Energy, both independent power producers with substantial nuclear portfolios, are the primary beneficiaries. In the second quarter alone, Constellation signed roughly 920 megawatts of long-term nuclear contracts with corporate customers, averaging 18.5 years and locking up about 30% of its clean baseload output under long-term agreements. The company also signed a power purchase agreement with Walmart for approximately 176 megawatts of wholesale supply from the Dresden Clean Energy Center across two 15-year terms starting in 2029 and 2030, building on earlier deals with Microsoft and Meta Platforms.
Vistra has moved even more aggressively. The company signed a massive power purchase agreement with Meta Platforms for 2,600 megawatts of energy and capacity at its PJM nuclear site and a long-term contract with Amazon Web Services for up to 1,200 megawatts from its nuclear plant in Texas. In June, Vistra partnered with KKR, Nvidia, and the Kuwait Investment Authority to form Helix Digital Infrastructure, an infrastructure development and financing company with $10 billion in capital commitments, including $1 billion from Vistra, with Vistra serving as the preferred power partner for both new-build and existing projects.
These multi-decade, fixed-price agreements with built-in inflation escalators provide both companies with highly predictable, long-duration revenue streams—a fundamental departure from the cyclical energy business model of the past. The preference for nuclear and low-carbon power sources reflects the fact that major hyperscalers have decarbonization mandates and require reliable baseload power that nuclear utilities can deliver.
Yet both stocks carry material downside risks. Constellation and Vistra face potential regulatory scrutiny around co-location and behind-the-meter deals, which could constrain future contract growth. More fundamentally, both companies depend on continued heavy AI capex; a slowdown in data center spending would directly reduce long-term energy demand projections. Currently, both stocks are down 32% from their 52-week highs, suggesting the market is weighing these risks heavily against the structural tailwind of rising power demand.
The energy industry is undergoing a fundamental shift driven by AI infrastructure expansion. Hyperscalers require massive amounts of continuous, reliable power to run data centers, but grid development is a long-lead project—taking four to 10 years or longer—while data center construction takes only up to two years. This structural mismatch has created a bottleneck in power availability that benefits established nuclear utilities with existing capacity and regulatory approvals.
Constellation Energy and Vistra Energy are capitalizing on this shortage by locking in long-term, fixed-price agreements with built-in inflation escalators, transforming what were once stable, low-growth businesses into high-growth AI-linked plays. Both companies have secured agreements with multiple hyperscalers: Constellation with Microsoft, Meta Platforms, and Walmart; Vistra with Meta Platforms and Amazon Web Services. The decarbonization mandates many tech companies operate under further favor nuclear and low-carbon suppliers, structurally supporting the economics of these utilities over fossil fuel alternatives.
However, both stocks face material risks. Regulatory scrutiny around co-location and behind-the-meter deals could constrain future growth, and a slowdown in AI capex would directly reduce projected energy demand increases. Both Constellation and Vistra are trading down 32% from their 52-week highs, reflecting these headwinds and investor caution despite the fundamental tailwind of rising data center power needs.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
The AI news that matters, in one minute each morning.
Sign up free