Artificial intelligence meets natural stupidity: Managing the risks
A potential electricity bubble may be created due to speculative AI demand forecasts, warns @AmoryLovins in this essay. Despite massive investments in AI data centers and predictions of soaring electricity needs, actual U.S. electricity consumption fell in 2023 and rose only 2% in 2024, with data centers accounting for just 4.5% of total use. Lovins, co-founder of the @RockyMtnInst, and @Stanford a Precourt Institute for Energy scholar and lecturer in the Department of Civil & Environmental Engineering, draws parallels to 1999 when the coal industry wrongly predicted information technology would consume half the nation's electricity, leading to costly overbuilding of unnecessary power plants.
Current policy responses include declaring energy emergencies to fast-track gas, coal, and nuclear plants that can't compete in markets, while stalling renewable projects that represent 93% of new U.S. capacity additions. The International Energy Agency expects renewables to increase electricity supply 10-20 times more than data centers increase demand.
Lovins recommends market-based risk allocation, requiring developers to guarantee power payments, and utilizing "power couples" ā combining clean energy with flexible data center demand ā to avoid repeating past overbuilding mistakes while supporting genuine AI growth.
Access the abstract and full essay here:
integrative-design-for-radical-energy-efficiency.stanfordā¦
#datacenters #electricity