Casey Crownhart and Pilita Clark frame a debate around a simple premise: in the era of Artificial Intelligence, the biggest barrier to progress is energy. In the United States demand from data centers is rising as billions of queries hit popular models, and efficiency gains no longer offset that growth. The result is higher electricity bills in regions where data centers strain the grid and a shortfall in new, steady power capacity to serve planned facilities.
China’s rapid expansion in power generation highlights the contrast. In 2024 China added 429 gigawatts of new capacity, more than six times the net capacity the US added in the same period. China is installing solar, wind, nuclear, and gas at record rates even as coal’s share declines. By comparison the United States is described as reviving coal, with aging coal plants running less reliably than before, and renewable projects facing political headwinds. The article also notes China already earns more from exporting renewables than the US does from oil and gas exports.
The conversation outlines short and longer term responses. Short term fixes include greater flexibility from data centers, such as curtailing consumption during grid stress. A Duke University study cited found that curtailing just 0.25 percent of the time could free roughly 76 gigawatts of capacity. Regulators are considering rules forcing tech firms to match their demand with generation and deals that let utilities use backup generators. But flexibility alone will not meet projected demand, which forecasts show could be less than double to four times today’s needs within five years. The piece calls for more renewable permitting, realistic planning for natural gas deployment, better public data on energy consumption by Artificial Intelligence systems, and caution about relying on unproven claims that Artificial Intelligence will solve the climate transition.
