Op-ed: data center backlash may save AI firms from overbuild
TL;DR
- Gartner projects the sector will spend $7 trillion on infrastructure through 2030 while generating only $3.6 trillion in related revenue over the same period.
- Voters rejected data centers in 40 of 41 local ballot measures, and analyst Paul Kedrosky puts 41% of planned capacity in states backing a pause.
- Anthropic has trimmed its IPO valuation band from $2 trillion toward the $1.3-$1.5 trillion range as investors re-price AI capex.
The $3 trillion AI-stock wipeout this month was less about fear of a regulatory crackdown than about investors finally pricing in the gap between what the industry spends and what it earns, Mark MacCarthy of Georgetown and David Klaus, a retired Deputy Under Secretary at the US Department of Energy, argue in Tech Policy Press. They anchor the piece on Gartner's estimate that the sector will "spend $7 trillion on infrastructure through 2030 but will generate only $3.6 trillion in related revenue" over the same period.
Running alongside the arithmetic is a political fact. Voters, the authors write, rejected data centers in "40 out of 41 local ballot measures." Analyst Paul Kedrosky, whom they cite, puts 41% of planned data center capacity in states now backing a pause.
Their counter-intuitive read is that this pushback is doing developers a favor. The backlash, in Kedrosky's phrase, "might save them from themselves" by preventing excess capacity the revenue cannot support. Instead of racing to pour more concrete for training runs, the authors argue, the industry should tilt spend toward inference, where customers actually pay.
They stack demand-side evidence that the frenzy is getting ahead of adoption. Goldman Sachs, they note, "has prohibited non-engineers from using advanced AI tools." Bank of England data pegs AI's contribution at "about 0.1–0.2 percentage points to annual productivity growth since early 2024." And Anthropic has trimmed the IPO valuation under discussion from around $2 trillion toward the $1.3-$1.5 trillion range, a move corroborated by Bloomberg's reporting on the company's $64.4 billion raise.
The piece is, in effect, an argument against the "Democrats killed the rally" story that surfaced after the midterms. In the authors' framing, the sell-off started weeks earlier, when the numbers stopped adding up.
Shared on Bluesky by 2 AI experts
Originally reported by techpolicy.press
Read the original article →Original headline: The Coming Data Center Slowdown Might be Good for AI Developers