NRG, VST, CEG: AI Power Demand as the Binding Constraint
US datacenter electricity demand to grow 200-300 TWh by 2030. NRG at 0.70x P/S, VST 2.11x, CEG 2.58x — three options for AI electricity thesis.
Three independent utility-equity stories converged on June 3. NRG Energy held its mid-2026 outlook on AI-related power demand acceleration. Vistra (VST) reaffirmed nuclear power purchase agreements with hyperscaler customers under negotiation. Constellation Energy (CEG) continued discussions on Three Mile Island restart with Microsoft as anchor customer. The thesis is simple: hyperscaler capex announcements add up to $400-500 billion through 2026, and that capex requires power. NRG, VST, and CEG are the three publicly-traded power producers with the cleanest expressions of "AI electricity demand growth" — but their stock-market positioning has diverged meaningfully through 2026, suggesting investors haven't yet settled on which name captures the cleanest bet.
Why the power-utility AI thesis exists
US datacenter electricity consumption is projected to grow from approximately 200 TWh in 2024 to 400-500 TWh by 2030 — adding 200-300 TWh of new demand against a US generation base of approximately 4,200 TWh. That's a 5-7% absolute increase in total US electricity demand from a single end-use category. Compared with the 1-2% annual growth that utilities had planned for over the prior decade, this represents an extraordinary demand-side regime change.
Three implications for power producers:
- Capacity factor improvement. Existing power assets that run below 100% capacity factor can capture incremental demand without new capex. Nuclear and natural gas baseload assets benefit most.
- Long-dated PPA pricing power. Hyperscalers signing 10-15 year power purchase agreements at fixed prices materially de-risk producer cash flows. CEG and VST have both announced multi-GW commitments.
- Behind-the-meter co-location. AI datacenters located at power plant sites bypass transmission constraints. This is the Three Mile Island restart strategy, complementing the grid equipment supply bottleneck that CEG and Microsoft have pioneered.
Data points
drillr terminal snapshot (June 3, 2026):
| Metric | NRG | VST | CEG |
|---|---|---|---|
| Market cap | $28.2B | $51.8B | $96.5B |
| June 3 close | $133.76 | $153.80 | $267.24 |
| Forward P/S | 0.70× | 2.11× | 2.58× |
| Forward EV/Sales | 1.21× | 2.99× | 3.16× |
| Forward revenue growth | +24.2% | +51.8% | +25.3% |
| EBITDA margin (TTM) | 9.3% | 34.1% | 23.6% |
| FCF margin (TTM) | -1.1% | 6.0% | 3.8% |
| FY25 revenue | $30.7B | $17.0B | $11.1B |
| FY25 operating income | $1.85B | $1.34B | $2.84B |
| Q1 2026 revenue | $10.3B | $4.65B | n/a |
| Q1 2026 operating income | $257M | $527M | n/a |
| Dividend yield | 1.37% | 0.15% | 0.61% |
| YTD price return | -16.2% | -2.1% | -22.8% |
| 1-year price return | -17.1% | -10.2% | -12.9% |
The valuation dispersion is the story. NRG at 0.70× forward P/S is the cheapest power name in the listed-utility complex — reflecting a more competitive retail electricity business mix and lower-margin generation portfolio. VST at 2.11× forward P/S has materially elevated multiples because of its nuclear and natural gas baseload assets with direct AI-customer commitments. CEG at 2.58× sits at the top of the cohort, reflecting the Microsoft Three Mile Island partnership and similar premium-pricing-power asset base.
The negative YTD returns across all three names are surprising given the bull thesis. The reason: rates exposure. Power utilities trade partly on long-term Treasury yields (PPA cash flows have bond-like characteristics), and the recent run-up in 10-year yields toward 4.5% has compressed utility multiples broadly. The AI thesis isn't fully reflected in equity prices yet — meaning the upside path requires both rates stability and continued AI demand growth.
VST's forward revenue growth of +51.8% is striking — it implies meaningful 2026 contract starts that haven't been fully reported. The company has been most aggressive in monetizing nuclear assets with hyperscaler customers under long-dated PPAs.
{
"hint": "A clean infographic showing three power generation facilities (nuclear plant, gas plant, datacenter) connected by power transmission lines. Each facility labeled with the relevant ticker — nuclear plant shows 'CEG / VST nuclear baseload', gas plant shows 'NRG natural gas + retail', datacenter shows 'MSFT / GOOGL / AMZN / META AI loads'. A small annotation reads '200-300 TWh new US demand by 2030'. Muted business color palette: soft blue, gold, dark green. Editorial publication aesthetic, no decorative elements.",
"aspect": "16:9",
"style": "minimalist editorial infographic",
"alt": "AI power demand thesis showing NRG VST CEG power generation supplying hyperscaler datacenter loads with 200-300 TWh demand growth by 2030",
"caption": "AI electricity demand — three utility plays positioned at the asset level"
}
Analysis: which one captures the cleanest exposure
The trade-off is risk-reward across the cohort.
CEG is the highest-quality but most fully-priced. Microsoft-anchored Three Mile Island restart provides validated premium pricing on nuclear capacity. Forward P/S of 2.58× is the highest in the cohort. Upside requires CEG to expand the model to additional reactor sites — possible but capacity-constrained. Implied 12-month total return: 5-15% plus dividend.
VST is the highest-growth but most leveraged. Multi-GW nuclear and gas commitments under negotiation provide visible revenue acceleration. Forward revenue growth of +51.8% is the cohort's highest. Forward P/S of 2.11× sits below CEG. If contracts close, VST re-rates toward CEG's multiple — implying 25-35% upside. If contracts slip, VST retraces.
NRG is the deep-value option but lowest pure-play AI exposure. Forward P/S of 0.70× — far below the cohort. Retail electricity mix dampens the pure AI demand exposure. Upside requires either AI-thesis multiple expansion or capital allocation actions (buybacks, special dividends). Implied 12-month total return: 10-20% plus dividend.
For a balanced cohort allocation, the asymmetric trade is: 50% CEG (quality anchor), 30% VST (growth optionality), 20% NRG (value defense). This positioning captures the AI-thesis upside while preserving quality and defense if the AI capex trajectory moderates.
What to watch
- VST quarterly nuclear/PPA contract announcements: Watch for incremental hyperscaler PPA signings or expansion announcements. Multi-GW deals materially shift the equity narrative.
- CEG Three Mile Island restart timeline: Public utility commission decisions and operational milestones. Restart on schedule sustains CEG premium; delays compress.
- NRG strategic actions: Watch for buyback acceleration, asset sale announcements, or strategic merger activity. Any capital return acceleration unlocks deep value.
- Hyperscaler quarterly capex commentary: MSFT (mid-July), GOOGL (late July), META (late July), AMZN (early August). Direct read on power-purchase agreement velocity.
- 10-year Treasury yield: Movement above 4.75% pressures utility multiples broadly; sustained below 4.25% supports re-rating.
- Independent grid operator AI demand forecasts (PJM, ERCOT, MISO): Updated load forecasts directly inform the capex demand thesis.
Want deeper analysis?
Ask drillr anything about NRG, VST, CEG — powered by SEC filings, earnings calls, and real-time data.
Try drillr.ai for freeRelated Research
Drillr can make mistakes. Information only — not investment advice. Learn more