In a new presentation, economist Jim Rickards, a former advisor to the CIA and the Pentagon, argues that the AI boom's real constraint is raw electrical power, not faster chips. Rickards states that "the sky high valuations of AI companies assume AI growth will be exponential," resting on the belief "that there is essentially infinite demand for AI products and that stock prices will go up forever." He argues these assumptions only hold if nothing impedes endless scaling. However, Rickards notes that each incremental gain in AI capability demands exponentially more compute, data centers, and electricity. "We are reaching the physical limits of AI too," he says, concluding that "the power production needed for AI won't scale." He asserts that "AI progress is beginning to slow" and "the limits of this technology and the financial problems with AI companies are becoming impossible to hide." Rickards believes "the market is beginning to catch on to that fact," framing his argument that a story built on infinite scaling cannot survive contact with a hard physical ceiling.