Correction or not, AI winners and losers will become clearer in 2026
For much of 2025 the big question for tech investors was how much more to buy. Now we are entering a phase of intense ROI appraisal, leading to more dispersion between AI winners and losers. Active stock selection will likely be key.
Authors
Widespread fear of an AI bubble means investors are increasingly interrogating companies’ returns on investment (ROI) in relation to AI. This will intensify in the months ahead, bringing volatility and divergence. Both offer opportunities.
Markets are already becoming more discerning. They are rewarding firms with visible monetisation (for example, Google’s cloud business) while questioning those where returns are less clear or less convincing (as seen in the wake of Oracle’s December earnings report).
It is tempting to treat “AI risk” as a single category. In practice companies face vastly different competitive pressures, business models and financing needs. A major stumble by a leading large language model (LLM) such as ChatGPT or Anthropic – whether due to competition or funding constraints – would undoubtedly hit sector sentiment and pull valuations lower across the piece. But AI revenues do not all accumulate in the same place, and monetisation can be dispersed and hidden. That complexity makes misattribution of both risk and value unusually likely.
Increasing divergence between individual companies, even where they operate in apparently similar parts of the AI spectrum, is a trend we expect to continue. There are several reasons for this.
1.Revenues don’t always emerge where they are expected
AI is used in many ways with very different revenue implications. Some users pay for AI directly through subscriptions or licences, while others access AI-enhanced tools without directly paying for the AI component. Many businesses are deploying AI behind the scenes to protect market share, drive conversion or improve unit economics. In these cases, monetisation is concealed in broader revenues.
Any analysis of AI monetisation requires an examination of the full stack, covering the applications users interact with; the LLMs powering them; and the compute infrastructure underneath. Value flows down this stack: the costs for model access and compute capacity translate into revenue for LLM providers and hyperscalers, regardless of how users pay.
2.Sometimes revenues are clearly visible, as with LLMs and cloud providers…
The most obvious evidence comes from the LLM companies themselves, where developer usage, enterprise licences and consumer subscriptions are already generating substantial revenue. Combined revenues are expected to reach tens of billions of dollars within a few years – on a par with established software firms.
Similarly, hyperscale cloud providers are reporting accelerating growth driven by AI workloads. Management teams across AWS, Azure and Google Cloud consistently describe demand running ahead of capacity. These indicators suggest AI monetisation is taking hold.
3. …but sometimes revenues are hidden, yet driving better economics
A second layer of monetisation is far more diffuse. Digital platforms such as Meta and Google use AI not as a product to sell, but as a tool to enhance advertising performance and engagement.
The uplift is real but not labelled “AI revenue.” The same is true for companies operating in many sectors where AI is deployed to drive improved conversion and profitability. This hidden monetisation is already large, and frequently underappreciated.
Schroders’ Economics Team has developed twin model scenarios exploring how an “AI boom” or “AI bust” might unfold. Both scenarios present potential difficulties for investors and economies. In the “AI Bust” scenario, for example, modelled on the market collapse following the bursting of the 1999-2000 technology bubble, a capex slump could trigger a mild recession and two years of stagnation.
That work explores some of the wider, longer-term unknowns associated with this transformative technology. For now, given the solid economic background (particularly in the US), markets may well continue their upward march.
The angst around AI ROI is real and will undoubtedly lead to more volatility in markets in 2026. As in previous innovation cycles, several AI-related firms, both large and small, could fail. But the revenues are emerging. It will take more than a few disappointments to undermine AI’s long-term potential.
Any references to sectors/stocks/securities are for illustrative purposes only and are not recommendations to buy or sell. Past performance is not a guide to the future and may not be repeated.
Subscreva o nosso conteúdo
Visite o nosso centro de preferências e escolha que informação deseja receber por parte da Schroders.
Authors
Topics