The AI boom's biggest bottleneck by 2026 wasn't chips or capital — it was electricity, and that scarcity turned traditional utility and power-generation companies into some of the least obvious, best-performing plays on the entire AI trade: Constellation Energy signed a 20-year deal with Microsoft to restart Three Mile Island specifically to power AI data centers, while Vistra spent $4.7 billion buying natural gas generation capacity to meet the same demand. This closes out the series by looking at the physical infrastructure layer underneath every chip and every dollar of capex covered so far — the electricity that actually has to flow before any of it works. This is reporting on disclosed deals and company statements, not investment advice.
Constellation Energy: The Clearest Single-Stock Nuclear Trade
Constellation Energy is the largest nuclear power producer in the United States, and it became the single clearest expression of what analysts have called a nuclear renaissance driven by AI when it signed a landmark 20-year power purchase agreement with Microsoft to restart the Three Mile Island nuclear plant specifically to supply AI data centers — a deal one industry analysis described as validating the entire nuclear-for-AI trade in a single announcement. The company has committed roughly $5.1 billion in capital expenditure to secure nuclear fuel and build inventory, and its quarterly dividend has climbed steadily from $0.141 per share in 2022 to $0.4265 by 2026, with management targeting continued 10% annual dividend growth. In July 2026, Constellation also invested in Blue Energy as demand from AI data centers kept surging, adding to its position across the nuclear supply chain rather than just its own generation fleet.
Vistra and Talen: Betting on Speed and Existing Capacity
Where Constellation leans into nuclear specifically, Vistra and Talen have taken more diversified approaches to the same underlying demand. Vistra announced a $4.7 billion acquisition of Cogentrix Energy in early January 2026, adding roughly 5,500 megawatts of modern natural gas-fired generation across key U.S. power markets including PJM, ISO New England, and ERCOT — a deal the company stated was designed directly to address surging demand from AI data centers and large industrial customers, expected to be immediately accretive with mid-single-digit free cash flow per share growth projected over several years. Talen Energy took a nuclear-specific route similar to Constellation's, partnering with Amazon in Pennsylvania and expanding that relationship in June 2026 to supply carbon-free nuclear power from its Susquehanna plant directly to AWS data centers in the region.
GE Vernova: A Different Point in the Value Chain
Not every AI-power investment story is about companies that generate and sell electricity directly. GE Vernova occupies a genuinely different position: it builds the turbines and grid infrastructure that enable power delivery in the first place, rather than selling electricity itself. That distinction shapes how each type of company experiences the same underlying boom — Vistra, Constellation, and Talen see demand show up first in power pricing and long-term contracts, while GE Vernova sees it show up in equipment orders and a growing multi-year backlog, giving it a different, arguably more forward-visible kind of demand signal than the generators themselves.
The Risk Investors Aren't Fully Pricing
The scale of underlying demand is genuinely enormous — data center electricity demand is projected to double or triple by 2030, straining a grid that industry analysts note was designed decades ago for residential and commercial load, not industrial-scale continuous computing. But some of the deals underpinning this trade carry real, underappreciated execution risk. Google has signed agreements for small modular reactor (SMR) capacity that won't actually be built until the late 2020s at the earliest — meaning the company is contracting today for generation capacity that doesn't physically exist yet, a forward commitment rather than an operating asset. Amazon has committed more than $100 billion to data center infrastructure, with each individual facility requiring dedicated utility-scale power secured before a single server can go live, meaning the pace of AI infrastructure buildout is now directly gated by how quickly power generation and grid capacity can actually be built, not just by chip supply or capital availability.
AI Power Plays at a Glance
| Company | AI power angle | Notable deal |
|---|---|---|
| Constellation Energy | Largest US nuclear producer | 20-year Microsoft PPA to restart Three Mile Island |
| Vistra | Natural gas and nuclear generation | $4.7B Cogentrix acquisition, +5,500MW gas capacity |
| Talen Energy | Nuclear generation, direct hyperscaler ties | Expanded Amazon/AWS deal at Susquehanna nuclear plant |
| GE Vernova | Power infrastructure equipment, not generation | Multi-year turbine and grid equipment order backlog |
Frequently Asked Questions
Why are utility stocks tied to the AI boom?
AI data centers require enormous, continuous amounts of electricity, and the existing power grid wasn't built for that scale of industrial demand. Utility and power-generation companies with reliable capacity, especially nuclear, have become key suppliers to hyperscalers, driving major long-term contracts and capital investment.
What did Constellation Energy's Microsoft deal actually involve?
Constellation signed a 20-year power purchase agreement with Microsoft to restart the Three Mile Island nuclear plant specifically to supply electricity for AI data centers, a deal widely described as validating the broader nuclear-for-AI investment trend.
Is there real risk in these AI power deals?
Yes. Some commitments, like Google's agreements for small modular reactor capacity, involve contracting for generation that hasn't been built yet and won't be operational until at least the late 2020s, meaning the actual power delivery timeline carries genuine execution uncertainty.
Is this article investment advice?
No. This is a factual summary of publicly reported utility and power company deals tied to AI data center demand, not investment advice. Anyone considering an investment decision should consult primary financial filings and a licensed financial advisor.