Goldman Sachs: AI Capex Could Reach $1.4 Trillion by 2027, Far Exceeding Wall Street Consensus

Claire Weston
Published 2026-06-11About 6 min read

Wall Street consensus puts hyperscaler capex at $757 billion this year, but Goldman Sachs argues actual 2027 spending could exceed consensus by more than 50% — implying systematic earnings upgrades across chips, memory, and data-center supply chains.

01

What is the current consensus actually pricing in?

Consensus expects hyperscalers — Amazon, Microsoft, Google and peers running massive data centers — to spend $757 billion this year, up 84% year-on-year.
By 2027, the consensus figure rises to $920 billion, still a record, but growth slows sharply to 22%.
This means → the market has already priced in "AI costs a fortune," but is also betting that peak spending intensity is now, and the pace cools from here.
02

Why does Goldman say the consensus is still too low?

Goldman lays out two more aggressive scenarios, both anchored to historical infrastructure build-out cycles.
Scenario one: if AI infrastructure investment follows the pattern of railroads and automobiles — incremental investment reaching 2–3% of GDP — hyperscaler capex could approach $1.1 trillion by 2027.
Scenario two: using free cash flow plus investment-grade credit-market capacity as the ceiling, spending could reach roughly $1.4 trillion, more than 50% above the current consensus.
In plain terms = Goldman's argument is simple: every past infrastructure mega-cycle spent far more than early forecasts predicted, and AI has no reason to be the exception.
03

If Goldman is right, what does it mean for investors?

If spending reaches those levels, current earnings forecasts for the chip, memory, and data-center supply chain all face systematic upward revision.
This means → AI-linked stocks that already look "expensive" may need to be re-evaluated — their valuation anchor, the earnings estimate itself, could be moving higher.
This reflects a deeper disagreement: most of the market is pricing in "AI investment is about to decelerate," while Goldman is saying "the acceleration isn't over yet."

Content is for reference only, not financial advice.