Even under generous assumptions, Bain estimates the AI industry will come up $800 billion short. If Bain is right, the industry is hurtling toward a wall where power constraints, limited GPU availability, and capital bottlenecks converge. If training remains cost-bound and data-center inference runs into power ceilings, more of the workload shifts to the edge. Microsoft recently bumped its Wisconsin AI data-center spend to more than $7 billion. Stay On the Cutting Edge: Get the Tom's Hardware Newsletter Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.