AI demand meets grid capacity

For the past 15 years, the default assumption about cloud computing has been simple: When enterprises needed more capacity, cloud providers would deliver it. The instance type might be scarce in one region. Maybe procurement would complain.