AI and the capex conundrum

Results season has brought a fresh wave of AI enthusiasm but also some concerns about margins, with the key ‘hyperscalers’ continuing to ramp up their capital expenditures in a familiar fashion to previous quarters.

This trend relates to those firms which are operating data centres on a massive scale, providing high volume networking services, in addition to storage and cloud computing, to their clients.

Much of this increase was justified by very strong operational performance, with Microsoft and Alphabet increasing their cloud revenues by 40 and 34 per cent, respectively, while Meta’s core advertising business grew 26 per cent.

This type of performance has been an integral driver of AI capex spending, with hyperscaler operating cash flows accounting for 60 per cent of AI spend.

Given the strong performance of the hyperscalers, this has given investors renewed faith that the beneficiaries of all this spending will continue to perform well.

Fresh perspectives

The numbers themselves certainly support this thesis, with Meta, Microsoft, Alphabet and Amazon expected to spend north of $480bn in capex during the US fiscal year ending in September 2026, a number which many expect to ramp up further as we close out the year.

This is obviously a huge absolute figure, but a closer look at how it’s funded by operating cash flows provides a fresh perspective. Over the past three years, capital intensity — capex as a proportion of operating cash flow — has steadily increased across the group. The most striking change comes from Meta, where this ratio is expected to rise from around 40 to more than 80 per cent in FY26.

The scale of this commitment sent the shares down more than 11 per cent post-report, with many questioning the end goal of the AI spend and likening it to Mark Zuckerberg’s 2022 pursuit of the ‘metaverse’.

These concerns are not without merit; the growing weight of spending is now starting to feed through into weaker free cash flow trends.

Meta’s free cash flow is expected to decline, in contrast to the anticipated increases for Alphabet, Microsoft and Amazon in FY26.

Contrast this with Alphabet, Microsoft and Amazon, whose cloud businesses are only limited by capacity and remain clear candidates for further investment, and it becomes clear why investors have shifted from Meta into the other names.

The next question is whether this combination of a slowing core business and increased investor scrutiny of spending is the first signal that the capex growth trajectory will slow.

Given the importance of companies like Meta in the AI funding ecosystem, any slowdown would likely provide an early indication of a pause in the AI trade and add to valuation concerns.

The response during this results season marks a clear shift in mindset, with investors no longer viewing all capex as equal. Even with strong core businesses, the rising capital intensity is beginning to weigh on sentiment where no clear roadmap exists.

While it is unlikely that the AI spending frenzy will end in the near term, this shift in sentiment shows how the investment community might force the hyperscalers to rationalise their spending.

As a result, the continued performance of the Magnificent Seven looks vital for AI momentum to continue and should be watched closely for any indications of an incoming demand air pocket. Even though many leading managers currently remain underweight the Magnificent Seven, their impact on investment returns is impossible to ignore.

Tom Dalrymple, global equity analyst at Aubrey Capital Management