Big Tech’s AI push faces a $1 trillion profit test, says UBS

The world’s six largest cloud companies would need to generate an additional $1 trillion in operating profit by 2031 to earn a 15% return on their investments, according to UBS Global Research, highlighting the scale of the returns required to justify the AI spending boom.

The six hyperscalers — Amazon, Microsoft, Alphabet, Oracle, Meta, and SpaceX — are expected to invest $7.7 trillion between 2026 and 2030. About $6.8 trillion of this would be additional spending compared with maintaining their pre-AI investment levels, UBS analysts said in a media note.

The additional profit required would be slightly less than twice the group’s current combined operating profit, the analysts estimated. At current cloud profit margins, earning that amount would require about $3.4 trillion in additional revenue.


UBS said achieving the targeted return would depend on AI delivering productivity gains and the companies funding its infrastructure capturing a meaningful share of the value created.

A tougher return equation

The spending marks a shift for companies that built highly profitable software, social media and ecommerce businesses without investing as heavily in physical infrastructure, according to UBS.

Those businesses have historically generated pretax returns on invested capital of more than 40%, the bank said. This measure compares the profit a business generates with the capital invested in it.

UBS expects returns on invested capital to fall by 20-25 percentage points as AI infrastructure accounts for a larger share of these companies’ businesses.

Who captures AI’s gains?

UBS estimates that the global economy, worth around $110 trillion, could reach $136 trillion by 2031 if it maintains the growth trajectory of the past decade, even without an additional boost from AI.

If AI raises annual productivity growth by 0.5%, 1% or 1.5%, global GDP could be another $3 trillion, $7 trillion or $10 trillion higher, respectively, by 2031, the bank estimated.

The challenge, UBS said, is converting a share of that additional economic output into revenue and profits for the companies investing in AI.

The bank compared the investment cycle with earlier spending booms in telecom, dotcom and cable infrastructure. Those investments helped improve productivity across the economy, but much of the value flowed to internet, ecommerce, social media, streaming and smartphone businesses, according to UBS.

Hyperscalers have a better chance of capturing the gains because they can earn revenue from AI through several parts of their businesses, the bank said.

Credit stress and power constraints

The questions around AI returns come amid a sell-off in global technology stocks and emerging markets, with weakness in bonds issued by some technology companies, particularly neocloud providers such as CoreWeave, emerging as one of the triggers.

However, UBS said the credit stress was concentrated among companies seen as having weaker balance sheets or large contingent liabilities. Higher US Treasury yields also explain almost all of the increase in hyperscaler bond yields this year.

Neocloud companies account for only 5-7% of expected AI data centre capital expenditure over the next four to five years, making them a relatively small part of the overall investment cycle, UBS said.

Despite the volatility, UBS said the cycle of corporate earnings upgrades across emerging markets and Asia remained intact, suggesting investor sentiment and fund flows were driving the sell-off more than weakening business performance. It identified comments from hyperscalers and developments in South Korea’s memory chip industry as factors to watch.

Power supply could also constrain the expansion, UBS said. Securing enough electricity for the planned data centres would depend on timely capacity additions, supportive legislation and smooth execution.