Correction or not, AI’s winners and losers will be defined in 2026
29 JAN, 2026
By Schroders
By Alex Tedder, Head of Equities at Schroders
For much of 2025, the key question for technology investors was how much more to buy. Now, we are entering a phase of intense scrutiny of return on investment (ROI), which will lead to greater dispersion between AI winners and losers. Active stock selection is likely to be crucial.
Widespread fears of an AI bubble are prompting investors to increasingly question whether companies’ AI investments are delivering sufficient returns. This scrutiny is set to intensify in the coming months, driving higher volatility and wider divergences – both of which can create opportunities.
Markets are already becoming more demanding. They are rewarding companies with visible monetization (for example, Google’s cloud business) while questioning those with less clear or less convincing returns, as seen following Oracle’s earnings report in December.
It is tempting to treat “AI risk” as a single category. In reality, companies face very different competitive pressures, business models, and funding requirements. A major stumble by a leading large language model (LLM), such as ChatGPT or Anthropic – whether due to competition or funding constraints – would undoubtedly hurt sector sentiment and push valuations lower across the board.
However, AI revenues are not concentrated in one place. Monetization may be distributed and often hidden. This complexity makes misattribution of both risk and value highly likely.
The growing divergence between companies – even those operating in seemingly similar areas of the AI spectrum – is a trend we expect to continue, for several reasons.
1. Revenues are not always generated where expected
AI is used in many different ways, each with very different revenue implications. Some users pay directly for AI through subscriptions or licenses, while others access AI-enhanced tools without paying explicitly for the AI component. Many companies deploy AI behind the scenes to defend market share, boost conversion, or improve unit economics. In these cases, monetization is embedded within broader revenue streams.
Any assessment of AI monetization requires analysis of the entire value chain – from user-facing applications, to the LLMs that power them, to the underlying computing infrastructure. Value flows along this chain: model access costs and computing capacity translate into revenue for LLM providers and hyperscalers, regardless of how end users pay.
2. Sometimes revenues are clearly visible, as with LLMs and cloud providers
The most obvious proof comes from LLM companies themselves, where developer usage, enterprise licensing, and consumer subscriptions are already generating substantial revenues. Combined revenues are expected to reach tens of billions of dollars within a few years, comparable to established software businesses.
Similarly, hyperscale cloud providers are seeing accelerated growth driven by AI workloads. Management teams at AWS, Azure, and Google Cloud consistently report demand exceeding capacity – clear signals that AI monetization is taking hold.
3. …but sometimes revenues are hidden, even if outcomes improve
A second layer of monetization is far more diffuse. Digital platforms such as Meta and Google use AI not as a product to sell, but as a tool to improve advertising performance and engagement.
The uplift is real, but it is not labeled as “AI revenue.” The same applies across many sectors where AI improves conversion rates and profitability. This hidden monetization is already significant and frequently underestimated.
Schroders’ economics team has developed two models exploring how an “AI boom” or an “AI bust” could unfold. Both scenarios present challenges for investors and economies. In the “AI bust” scenario – drawing parallels with the 1999–2000 tech bubble collapse – a decline in fixed investment could trigger a mild recession, followed by two years of stagnation.
This research highlights the broader and longer-term uncertainties surrounding this transformative technology. For now, given the strong economic backdrop, particularly in the United States, markets may well continue their upward trajectory.
Concerns around AI return on investment are real and will almost certainly drive greater market volatility in 2026. As with previous innovation cycles, some AI-related companies – large and small – are likely to fail. But revenues are emerging, and it will take more than a handful of disappointments to undermine the long-term potential of artificial intelligence.
