AI boom's hidden bill: Big Tech's spending hits $165 billion as free cash flow shrinks

The artificial intelligence boom is delivering spectacular earnings growth for US companies, but there is a less comfortable number emerging beneath the headline profits: free cash flow. 

The four major US hyperscalers — Microsoft, Amazon, Alphabet and Meta — are pouring unprecedented sums into AI infrastructure, according to a report by Jefferies’ Global Head of Equity Strategy Christopher Wood. Their combined capital expenditure reached $165 billion in the second quarter of 2026, up sharply from $72 billion in the first quarter of 2025.

At the same time, their aggregate free cash flow fell to just $7 billion in Q2 2026, from $60 billion in Q4 2025. Jefferies expects that figure to turn negative, at -$12 billion, in Q3 2026. 

MUST READ | The AI reality check: 85% of hyperscalers' cloud revenue is still non-AI

That creates a striking mismatch: AI is becoming increasingly important to corporate earnings, but building the infrastructure needed to power the AI economy is consuming huge amounts of cash. 

The spending is not necessarily a sign that the AI boom is failing. In fact, Jefferies argues that the AI capex race remains strongly earnings-accretive for the companies supplying the infrastructure.

Chipmakers, semiconductor equipment companies, server manufacturers and other "picks and shovels" players can recognise revenue and profits as the infrastructure is sold, while the hyperscalers have to absorb the capital cost and depreciate those assets over time. 

That distinction could become increasingly important for investors. 

The depreciation bill is rising 

The immediate cash outflow is only part of the story. As the hyperscalers build more data centres and buy more computing equipment, depreciation expenses will rise. 

DON'T MISS | AI may reshape Indian jobs, but physical work could offer some protection, Goldman Sachs says

The combined depreciation expenses of Microsoft, Amazon, Alphabet and Meta reached $44.5 billion in Q2 2026, up 24% year-on-year. 

The question for investors is therefore shifting from whether Big Tech can afford to build AI infrastructure to whether the eventual revenue generated from that infrastructure will justify the enormous capital investment. 

The report points to an important source of optimism: cloud revenues for Microsoft, Google and Amazon rose 38% year-on-year to $126 billion in Q2 2026. That growing cloud business is central to the argument that today's AI spending could eventually generate adequate returns. 

DO CHECKOUT | Meta's Mark Zuckerberg says AI creates jobs as data centres expands

But the scale of spending means the payoff will have to be substantial. 

For now, the AI boom remains earnings-accretive. Jefferies estimates that the S&P 500's AI basket could deliver annualised EPS growth of 48% during 2026-27, compared with 23% for the broader S&P 500 and 12% for the index excluding AI.