Goldman's Jan Hatzius: AI spend won't last forever

Goldman Sachs chief economist Jan Hatzius used the firm's Communacopia & Tech conference to push back on the idea that AI spending can rise indefinitely. His baseline is optimistic — AI spending is sustainable, productive and contributes to stronger productivity growth — but he still expects an eventual slowdown as the industry moves from an investment phase into an exploitation phase. In that second phase, the technology is used and investment volumes fall, which he says will create challenges and have negative effects on people who assumed the boom would go on forever.

That warning sits alongside projections that point the other way. PwC's new Global Data Centre Outlook puts global AI infrastructure investment at a record $31.6 trillion through 2050, with annual data centre capital expenditures forecast to rise from roughly $800 billion per year in 2026 to $1.8 trillion per year in 2050. PwC expects the investment to accelerate as chips and other internet-connected equipment need upgrades every few years, and its global infrastructure leader for Australia, Clara Cutajar, says AI infrastructure is becoming one of the defining capital allocation challenges of the next generation, cutting across technology, energy, real estate, supply chains, regulation and financing.

Companies from Meta to Google to Microsoft are investing billions to support their AI ambitions, which is the backdrop for Hatzius's caution. The two views are not strictly contradictory — PwC's numbers describe the scale of the build-out, while Hatzius describes its life cycle — but they frame the stakes differently. For infrastructure investors, the question is likely to be how much of today's spending survives the transition Hatzius describes, and whether the returns justify the capital requirements that PwC highlights.