Why We’re Overbuilding AI Data Centers (And What Happens Next)
A Texas Talks discussion examines whether the rapid expansion of AI data centers is creating an infrastructure bubble. Dr. Joshua Rhodes says Texas and the broader industry may overbuild computing capacity, leading to bankruptcies and consolidation, while electricity demand remains a major driver of investment.
UT Austin energy researcher Dr. Joshua Rhodes discusses the possibility that projected AI data center demand, including estimates of hundreds of gigawatts, is overstated. He compares the current buildout to the early railroad boom, when the United States built substantially more track than was ultimately needed, and predicts that some data center projects will fail or consolidate even though AI technology remains economically useful.
The conversation also focuses on the value of securing data center power capacity, which can generate hundreds of thousands of dollars annually per megawatt and incentivizes companies to move quickly. Rhodes says emerging technologies such as small modular nuclear reactors and geothermal power could eventually help meet demand, but they must overcome high first-of-a-kind costs, licensing requirements, safety concerns and reliability challenges before competing with established generation technologies in Texas.