Powering AI

Why It Matters

Without sufficient, affordable power the U.S. risks losing its AI leadership to rivals, while rising electricity costs could trigger public resistance and slow investment in critical data‑center infrastructure.

Key Takeaways

• AI model growth drives a 12% rise in commercial electricity demand
• US added 55 GW generation in 2025, versus China’s 543 GW
• Half of US coal plants date back to the Carter era
• Real electricity prices rose 21% from 2021 to 2024
• Diversified generation—solar, gas, wind, batteries—needed within five years

Pulse Analysis

The surge in AI model size and the proliferation of data‑center clusters are turning electricity from a peripheral concern into a strategic bottleneck. While energy accounts for only 2‑6% of AI training costs, the sheer scale of future inference workloads could push that share higher, especially as models run continuously for days. Data‑center operators can mitigate peaks through demand‑response and flexible grid services, but any sustained supply‑demand imbalance will directly curtail AI training cycles and slow innovation. Understanding this dynamic is essential for investors and policymakers monitoring the AI‑energy nexus.

Compared with China’s aggressive grid expansion—543 GW of new generation in 2025 and over 40,000 km of high‑voltage lines—the United States has added just 55 GW and under 1,500 mi of comparable transmission. An aging fleet of coal plants, decades‑old transformers, and transmission corridors built during the moon‑landing era further constrains capacity. These structural deficits have already driven real electricity prices up 21% since 2021, outpacing inflation and eroding the cost advantage of U.S. data‑center hubs. The price trajectory threatens both corporate margins and consumer sentiment, especially in low‑income and rural communities where electricity bills compete with gasoline and food expenses.

To safeguard AI leadership, the U.S. must pursue a multi‑pronged grid strategy. Short‑term solutions include rapid deployment of solar, natural‑gas peakers, on‑shore wind, and battery storage, while long‑term investments focus on replacing aging infrastructure and installing advanced transmission technologies such as HVDC and dynamic line rating. Streamlined permitting, community‑engagement frameworks, and rate designs that allocate costs to beneficiaries will accelerate build‑out and reduce political friction. By aligning energy policy with AI development goals, the United States can keep power costs manageable, maintain grid reliability, and preserve its competitive edge in the global AI race.