What’s going on with U.S. power prices & AI/data centers

Contrary to popular belief, the recent surge in U.S. electricity prices, up 6% year-to-date, is primarily driven by re-industrialization rather than AI infrastructure, which accounts for only about 3% of power demand, according to analysis from Thunder Said Energy and the R Street Institute presented at a Jefferies summit. Experts have revised U.S. load growth forecasts significantly downward to 1% annually due to efficiency gains in software, hardware, and load flexibility, suggesting that investment focus should shift towards grid efficiency, utilization, and low-voltage opportunities, rather than overbuilding traditional generation capacity.

Analysis

The recent 6% year-to-date surge in U.S. electricity prices is primarily driven by re-industrialization, not artificial intelligence infrastructure, according to analysis from Thunder Said Energy and the R Street Institute. AI currently accounts for only about 3% of U.S. power demand, challenging the widespread assumption that data centers are the main cause of rising power costs. U.S. load growth averaged 1.5% annually from 2019 to 2024, almost entirely due to industrial loads, with geographic price increases further supporting this narrative.

Efficiency gains in software and hardware are significantly mitigating AI's power footprint; for instance, Google Gemini uses approximately 18 watt-hours per AI query, less than a Google search five years ago. Rob West of Thunder Said Energy has consequently revised U.S. load growth forecasts downward to 1% per annum from 3%. Furthermore, AI can enhance efficiency in other sectors, as demonstrated by Freeport's AI-enabled copper concentrator, which uplifted production by 5-10% while reducing energy intensity.

This revised outlook suggests risks of overbuilding combined-cycle gas turbines, especially if projected load flexibility and efficiency gains materialize. Investment focus should shift towards grid efficiency, utilization, and low-voltage opportunities, rather than solely on traditional generation capacity. The market already recognizes these opportunities, with many power electronics companies performing well year-to-date.

Key Decisions for Investors

• Investors should re-evaluate long-term investment theses in traditional power generation, considering the downward revision of U.S. load growth forecasts and the primary driver being re-industrialization rather than AI.
• Focus investment strategies on companies specializing in grid efficiency, utilization technologies, and low-voltage solutions, as these areas are poised for growth and mitigate risks of overbuilding generation capacity.
• Monitor industrial sector growth and energy demand trends as a key indicator for future electricity consumption patterns, while also assessing AI's potential to drive efficiency gains across various industries.