AI pivots from rewarding broader theme to ‘selective’ returns

Yasmine Raso

Senior Journalist

28 April 2026

Investment in artificial intelligence (AI) has entered a new phase where returns have “diverged” between the winners and losers, as whole-of-theme gains have moved aside for a “selective”, individual company-based approach.

That is the latest assessment from investment manager, Schroders, which marks a strong shift in sentiment for investors as “AI-related stocks are no longer moving in tandem” and performance becomes increasingly tied to the ‘winners’.

“AI has entered a new phase, where broad exposure is no longer enough and returns are becoming far more selective,” Ben Arnold, Investment Director, Global Equities at Schroders, said.

“Markets are now moving to a stock-by-stock assessment of who is best positioned to deliver sustainable returns, rather than rewarding the theme as a whole.”

Arnold noted three key components driving this new phase of AI investment and “where value will emerge”: deployment of capital, debt and demand. He said investors are diving deeper than ever into not just how much technology companies are investing in AI infrastructure but also how effectively it is being used.

This comes at the same time as debt adds a layer of complexity and risk across the AI investment space, with companies becoming increasingly subject to judgement based on balance sheet strength and ability to sustain investment.

“A year ago, rising capital expenditure was seen as a sign of confidence and leadership,” he said.

“Tech businesses are committing vast sums to AI infrastructure; across chips, networking, data centres and cloud capacity.

“However, the market is increasingly questioning whether all this investment will generate sufficient returns.

“Take Oracle, issuing almost as much debt since January 2025 than in the previous seven years combined in a bid to accelerate its data centre build out.

“The perceived risk of its debt rose quickly in Q4 2025, signalling investor scepticism around its ability to catch up in the race for AI leadership and generate sufficient returns to justify both the investment, and the leverage used to fund it.

“The growing use of leverage is amplifying both opportunities and risks. But not all leverage is being treated equally, with markets more comfortable with the credit profiles and capital structures of other hyperscalers.”

Arnold indicated that demand is the “most difficult to factor to assess”, as strong AI take-up hasn’t necessarily “translated into immediate revenue”. Companies are now facing an entirely different approach from investors at the same time as having to balance “short-term monetisation with long-term strategic investment”.

“Understanding where real, durable demand sits requires much deeper analysis, as usage, pricing power and revenue can diverge significantly across the AI value chain,” Arnold said.

“Even at the individual company level, the link between demand and revenue can be unclear. This was evident in the market’s reaction to Microsoft’s recent earnings. Slower-than-expected growth in its cloud computing platform was initially seen as a leading indicator of weakening demand.

“However, management clarified this reflected a deliberate decision to redirect capacity towards internal AI development (such as Copilot), aimed at driving long-term monetisation.

“Broad exposure to the theme has worked up until recently, but it’s becoming clear that stock selection, not general thematic exposure, will drive the next leg of returns.

“Diversification remains critical, but the focus now is on identifying the companies that can execute and deliver sustainable returns through the cycle.”