Power-stock prices and power demand are different indicators
On July 28, data-center power and cooling suppliers Vertiv and Eaton and electricity-related Constellation Energy, Vistra and GE Vernova fell roughly 3% to 6%. The move did not follow an announcement that AI data-center electricity demand had suddenly contracted; it occurred during a broader rotation away from semiconductors and highly valued growth assets.
The IEA’s July 23 update expects global electricity demand to grow 3.6% in 2026 and 3.8% in 2027, after 3% in 2025. Industry, cooling, electric vehicles and data centers remain structural drivers, with U.S. consumption forecast to rise close to 2% in 2026.
Why shares can fall while demand rises
- Valuation and rates: Power equipment and generation assets are valued on long-duration cash flows. Higher long-term rates reduce their present value and raise construction financing costs.
- The gap between demand and revenue: Announced data-center gigawatts require land, substations, transmission, permits and construction before they become energized load, cooling orders and generation revenue. Delays can move revenue even when demand forecasts remain valid.
- Grid cost allocation: Profitability depends on whether interconnection and network upgrades are paid by the data center, utility, generators or existing customers.
- Supply and margin execution: Rapid expansion adds orders but raw materials, tariffs, labor, warranties and fixed-price contracts can pressure margin.
Medium-term outlook: execution quality matters more
EIA’s baseline had U.S. electric load growing 1.9% in 2026 and 2.5% in 2027. In a scenario where data-center regions grow 50% faster than baseline, EIA modeled ERCOT’s 2027 wholesale price at 79% above its baseline forecast. Demand is therefore both an investment opportunity and a price and reliability risk.
Key checkpoints are the duration, price and credit support of data-center power agreements; actual interconnection dates; equipment orders and book-to-bill; plant availability; transmission investment and regulatory approvals; and free cash flow. Operating capacity matters more than announced capacity, while backlog cancellation and margin matter more than backlog headlines alone.
The base case still supports continued AI-related investment in power, cooling and grids. Not every project will operate on schedule, however, and high capital costs and interconnection constraints can make equity prices much more volatile than electricity demand.
This analysis is based on public information and is not investment advice.