Updated
Updated · CNN · Aug 6
AI Data Centers Miss 50% of 2028 Capacity Targets as 60% of 2027 Projects Stall
Updated
Updated · CNN · Aug 6

AI Data Centers Miss 50% of 2028 Capacity Targets as 60% of 2027 Projects Stall

3 articles · Updated · CNN · Aug 6

Summary

  • Only about half of AI computing capacity slated to come online by 2028 is expected to meet its target date, far worse than the roughly 72% on-time rate for data centers historically, Goldman Sachs said.
  • JPMorgan said 60% of capacity planned for 2027 has not even started construction and another 7% of projects already underway have been delayed, despite $750 billion in AI infrastructure investment this year.
  • Shortages are piling up across the buildout: TSMC remains a key chip bottleneck, wait times for generation step-up transformers have tripled, and the US would need 500,000 more electricians plus hundreds of thousands of welders and plumbers.
  • Public resistance is adding friction, with 71% of Americans opposing AI data centers in a Gallup poll and temporary moratoriums recently enacted in New York and Texas, though Goldman says permitting is a bigger obstacle than outright bans.
  • Even so, Census data showed data center construction spending rose 7% in June to $68.3 billion, up 46% year over year, as developers pursue a boom that may still fall far short of nearly 4,000 announced US projects.

Insights

How will the hidden shortage of half a million construction workers quietly derail the ambitious future of American artificial intelligence?
Could the invisible wall of power grid delays ultimately trigger the collapse of the massive AI infrastructure bubble?
With transformer waits exceeding three years in 2026, will tech giants completely abandon traditional grids for private microgrids?