Wharton professor: hyperscalers need $1.1T AI spend and 2.7x productivity gain by 2030

Wharton professor Jessica Wachter estimates hyperscalers must spend about $1.1 trillion on AI infrastructure by 2027 and raise their own productivity 2.7 times by 2030 to break even. If these plans fail, the investment would become the largest inefficient capital allocation in history.

• Hyperscaler AI infrastructure spending in 2026 to reach about $750 billion
• AI revenue this year estimated at $150–200 billion
• Break-even by 2030 requires 2.7x productivity growth
• Alphabet free cash flow turned negative by $5.9 billion for first time since 2004