Big Tech’s capex arms race

Investors have grown accustomed to periodic leaps in capital spending by the largest US tech companies. But changes in Big Tech’s business raise serious questions about whether the returns from the latest spending binge will be as predictable as those of the past.

After jumping 32 per cent in 2020, the combined capital budgets of Alphabet, Amazon, Apple, Meta and Microsoft rose as much again in 2021, reaching $140bn. They climbed another 20 per cent in the first nine months of this year.

Periodic spurts like this are nothing new. One reason has been the need to add capacity ahead of expected surges in growth. New data centres, offices to house extra workers, or (in the case of Amazon) warehouses and delivery vans all need to be put in place ahead of time. The surprisingly strong growth rates that the big tech companies have maintained in recent years — particularly through the pandemic — have largely justified their confidence.

Yet they have not always got the timing right. Exhibit A is the spending binge that Amazon began shortly before Andy Jassy took over as chief executive in mid-2021. Capex of $61bn last year was a big leap from the $17bn of only two years before. As growth slowed earlier this year, it became clear that Amazon had put its foot on the accelerator too hard.

Going into an economic slowdown, this risk has increased — particularly since some corners of the digital economy are showing signs of reaching maturity. Big Tech’s latest earnings reports and forecasts pointed to an unexpectedly pronounced deceleration in ecommerce and digital advertising as consumer demand slows. Now that they represent such a large share of the overall economy, it is no longer so easy for the tech companies to shrug off the macro trends.

Meanwhile, even if demand for some services is not slowing down, the weakening economy is making customers less willing to pay premium prices.

This led to slower than expected growth in the cloud computing businesses of Amazon and Microsoft in the latest quarter. Both companies said customers had been “optimising” their cloud spending — in other words, moving to cheaper data storage plans or running their workloads on less expensive chips. Volumes were unaffected but revenue disappointed.

Yet the demand for extra computing power has been growing exponentially. The rise of cloud computing and, more recently, a push to enhance their artificial intelligence capabilities, has brought a step-change in Big Tech’s capital spending.

Meta, which disappointed Wall Street last week with plans to intensify its spending on the metaverse, also said that the demands of AI were causing a leap in capital spending. From 16 per cent in 2021, Meta’s capex is expected to reach 28 per cent cent of revenue this year, before hitting 30 per cent in 2023. Changes on this scale have raised concerns on Wall Street that the capital intensity of Big Tech’s business is changing.

In cases where the returns seem likely to match the higher spending levels, investors have taken this in stride. Amazon’s spending as a percentage of revenue jumped from around 5 per cent in 2016-2019 to reach more than 12 per cent in both 2021 and the first nine months of this year. New technology, mainly to support the expansion of Amazon Web Services, caused much of this. The push into cloud computing and advertising (another business driven by tech) has boosted Amazon’s profit margins.

Yet the likely returns from much of Big Tech’s spending are getting harder to predict. The recent efforts by customers to drive down their cloud bills suggest that the tech companies are likely to face pressure on pricing as the cloud becomes a bigger slice of overall IT.

At the same time, it is uncertain whether the promised improvements that AI will bring to existing services — or the entirely new services it will make possible — will produce enough new revenue to justify the cost.

Meta says AI will help it analyse all the content flowing across its network so that it can put the most appropriate images or videos in front of each user, while also targeting its advertising more precisely. The effectiveness of this is unproven.

In some cases, AI is even becoming the product. The generative AI systems that produce text or images on command have fuelled the latest race in the AI world to build ever-larger — and more expensive — models.

It is hard to tell how expensive this arms race will become, or what extra profits it will produce. But even as it races into a downturn, Big Tech is not about to let up on the spending.

richard.waters@ft.com