Morgan Stanley analyst Stephen Byrd increased his estimated US data center power gap through 2028 from 30 gigawatts into the mid-30s gigawatts, following gubernatorial outcomes in Texas and Ohio that added approval steps for developers.
The revised estimate of a larger power shortfall in the US suggests that the rapid expansion of AI infrastructure could face delays. Byrd frames this gap as a timing issue, pushing deployment into 2029 and potentially elongating the AI buildout cycle. He dismisses grid optimization as insufficient, stating utilities manage for worst-case scenarios and cannot easily allocate average spare capacity to data center tenants. This constraint highlights a significant infrastructure challenge for the burgeoning AI industry, requiring developers to consider alternative solutions like off-grid builds or explore markets in other regions, such as Louisiana and Asia.
Sources