What the split between AI as an investment theme and AI as a financial utility says about the next phase of retail investing
Research current through October 5, 2026
| Research Thesis
Retail investors are not abandoning artificial intelligence. They are separating two decisions that were once bundled together: whether AI is useful and whether AI-linked stocks are worth buying at today’s prices. Survey evidence shows AI tools have moved from novelty toward routine financial utility even as expectations for AI stocks have cooled. That decoupling matters because it suggests the AI trade is maturing from a broad narrative bet into a valuation-and-returns problem: investors can increasingly trust the tool while demanding much more proof from the companies financing the buildout. |
Key numbers
| Metric | What it measures | Source |
| 56% | Global retail investors who use or are open to AI for investment decisions in 2026 | eToro |
| 44% | Global retail investors expecting AI stocks to rise, down from 55% a year earlier | eToro |
| 67% | U.S. affluent/HNW investors worried about an AI bubble in the next 12 months | Janus Henderson |
| 61% | Same survey: expect AI to have a positive impact on returns over five years | Janus Henderson |
| $800B | Estimated 2026 hyperscaler capex, rising toward $1.1T in 2027 | Morgan Stanley |
The paradox is real — but it is not quite the headline version
The cleanest evidence comes from eToro’s September 2026 Retail Investor Beat. Across 11,000 investors in 13 countries, 56% said they already use or are open to using AI tools such as ChatGPT or AI agents to pick or alter investments. That is slightly below 58% a year earlier, but materially above the 45% recorded when eToro first asked the question in 2023. In other words, AI adoption has stopped looking like a fad even though it is no longer accelerating in a straight line.
At the same time, optimism toward AI stocks has weakened much more sharply. Only 44% of global respondents expect AI-related stocks to rise in 2026, down from 55% a year earlier, while the share expecting declines increased to 18% from 11%. In the United States, the same split is visible: 55% use or are open to AI for investing, while 47% expect AI stocks to rise, down from 57% a year ago.
That creates a useful original measure: the ‘utility-versus-equity gap.’ Globally, the gap between willingness to use AI and bullishness on AI stocks widened from 3 percentage points in 2025 to 12 points in 2026. In the U.S., it widened from 1 point to 8. The paradox is therefore less ‘trust rising while trust falls’ than a decoupling: confidence in AI as a tool has proved much stickier than confidence in AI as a trade.
Figure 1. Global retail AI-tool adoption versus expectations for AI-stock gains. Source: eToro Retail Investor Beat; calculations by Dave Finances.
Figure 2. The same utility-versus-equity gap in the United States. Source: eToro Retail Investor Beat; calculations by Dave Finances.
AI is moving from narrative to utility
That decoupling is what normally happens when a technology moves from story to infrastructure. During an early investment boom, the technology itself and the stocks associated with it are often treated as one idea. Later, users begin to consume the technology independently of whether they want to own the suppliers. Cloud computing is useful even when cloud software valuations are expensive. Smartphones can transform daily life without every handset maker producing superior shareholder returns. AI is beginning to enter the same stage.
Retail platforms are accelerating that transition. Schwab launched AI-generated portfolio insights that combine account performance, market news and in-house research. Robinhood’s Cortex can answer portfolio questions using live account data, and Robinhood has announced Agents designed to analyze markets, build strategies and trade on a customer’s behalf. SoFi’s Composer turns natural-language investment ideas into backtested, rules-based strategies, while eToro’s Tori can create and manage AI-driven Agent Portfolios conversationally.
This matters because the unit of adoption is changing. Investors are no longer merely buying an ‘AI stock.’ They are using AI to summarize earnings, screen securities, test rules, monitor portfolios and increasingly automate actions. As the technology disappears into the workflow, its value can rise even while the market’s willingness to pay a premium for AI-labelled equities falls.
Figure 3. AI investing is not a binary adoption decision: trust can differ sharply by level of agency. Dave Finances framework.
| Platform | 2026 AI capability | Agency level | Key guardrail / distinction |
| Schwab | Portfolio insights combining holdings, news and Schwab research | Explain | Informational; explicitly not investment advice |
| SoFi | Composer builds, backtests and automates natural-language strategies | Automate | Executes predefined rules rather than open-ended agent decisions |
| eToro | Tori can create and manage AI-driven Agent Portfolios | Suggest / manage | Conversational agent inside regulated investing platform |
| Robinhood | Cortex uses live account data; Agents announced for strategy building and trading | Delegate | Moves furthest toward autonomous action; user controls remain central |
Table 1. AI is moving deeper into the retail investing workflow, but platforms expose users to different levels of automation.
The behavioral-finance explanation: instrumental trust is different from valuation trust
Behaviorally, there is no contradiction in trusting an AI tool and distrusting an AI stock. The two decisions ask different questions. Tool adoption asks: does this make my research faster, clearer or cheaper? Stock selection asks: will this company earn enough incremental cash flow to justify its current valuation and capital spending? A user can answer yes to the first and no to the second.
Research on algorithmic decision-making helps explain why trust can be highly conditional. Work by Jennifer Logg, Julia Minson and Don Moore documented ‘algorithm appreciation’: lay participants often weighted algorithmic advice more heavily than human advice in forecasting tasks. But Berkeley Dietvorst, Joseph Simmons and Cade Massey documented ‘algorithm aversion’: once people see an algorithm make a mistake, they can abandon it faster than an imperfect human forecaster. Those findings are not mutually exclusive. They imply that trust depends on framing, experience, control and the consequences of error.
Current investor surveys look similar. Schwab found more than 60% of surveyed retail clients were interested in AI and nearly 70% believed it could play a meaningful role when paired with human expertise. Yet Vanguard’s September 2026 survey of more than 6,000 investors found 57% were uncomfortable allowing AI to make decisions on their behalf. The line appears to be agency: investors are more willing to let AI inform judgment than replace it.
Why AI stocks now face a much higher burden of proof
The other side of the paradox is financial rather than psychological. The AI buildout has become so capital-intensive that investors no longer need to doubt the technology to question the stocks. Morgan Stanley estimates combined hyperscaler capital expenditure at roughly $800 billion in 2026 and $1.1 trillion in 2027. Goldman Sachs separately estimates global AI investment will exceed $1 trillion in 2026 once spending beyond U.S. hyperscalers is included.
Those numbers change the question from ‘Will AI be important?’ to ‘Who earns an adequate return on this spending?’ This is why a retail investor can use an AI assistant every day and still sell a hyperscaler after a disappointing earnings call. Adoption proves demand for the technology; it does not prove that every layer of the supply chain captures enough economics to justify the investment required.
The September eToro survey captures that shift from promise to proof. Retail investors were still 2.5 times more likely to say Big Tech’s AI spending made them more likely to invest than less likely. But broader expectations for AI stocks cooled substantially. The bar is therefore rising inside the theme: investors increasingly want evidence of revenue conversion, margins, utilization and return on invested capital rather than more capex announcements.
Figure 4. Estimated combined hyperscaler capital expenditure. Source: Morgan Stanley Research.
| Investor Takeaway
The adoption question and the valuation question are diverging. More AI use can be bullish for the technology while simultaneously raising the amount of revenue that infrastructure owners must generate to justify the capital already committed. |
The AI trade is fragmenting into builders, owners and users
This shift also changes what ‘AI investing’ means. During the first phase, much of the market rewarded the builders: semiconductor companies, data-center operators, hyperscalers and model developers. In the next phase, value can migrate toward companies that use AI to lower costs, increase conversion, automate customer service or improve pricing without carrying the same infrastructure burden.
Morgan Stanley has explicitly framed the next phase around adoption rather than infrastructure alone. Its research argues that the investment opportunity may broaden from companies building AI capacity to businesses that integrate AI deeply enough to improve productivity and margins. That would make the investment theme less concentrated even as AI itself becomes more ubiquitous.
For retail investors, this is an important distinction. If AI becomes a general-purpose utility, the best AI investment may eventually be a company whose investment case is not marketed as ‘AI’ at all. The winners could be insurers with better underwriting, banks with lower service costs, retailers with better inventory planning or industrial companies with higher asset utilization. Utility spreads faster than narrative labels.
The platform layer may capture value before the user notices
There is another consequence: brokerage and fintech platforms can monetize AI adoption even if users become less enthusiastic about AI stocks. AI can increase engagement, reduce research friction, support premium subscriptions, generate more personalized prompts and potentially increase trading or assets under management. That creates a new revenue pool around the investment process itself.
Robinhood’s Cortex Digests had already been used by nearly one million customers by the first quarter of 2026, according to the company. Schwab is embedding AI directly into account summaries. SoFi is integrating AI into financial planning and systematic strategy design. eToro is repositioning its app around a proactive AI agent. None of those business cases requires Nvidia, Meta or another AI-linked stock to outperform the market next quarter.
This is why ‘AI as financial utility’ may be a more durable commercial story than ‘AI as the hottest sector.’ Platforms can benefit whenever AI reduces the cost of servicing, analyzing and engaging a customer—even if market leadership rotates away from the current AI winners.
But greater AI trust creates a new set of retail-investor risks
The utility thesis should not be confused with a claim that AI financial advice is mature or safe. The Financial Conduct Authority’s August 2026 research found that four in five less-experienced investors aged 18 to 40 had used AI for help with investing, and 56% said they trusted AI tools. But 44% incorrectly believed AI-generated financial information was regulated, 38% thought it was acceptable to make an investment decision solely from AI output, and 32% wrongly believed compensation schemes or the Financial Ombudsman would protect them if AI guidance went wrong.
That is a classic setup for automation bias: a tool becomes useful enough that users stop treating it as a tool. Large language models can summarize data rapidly, but they can also hallucinate, omit material context, misunderstand securities, overstate confidence or generate persuasive explanations for weak conclusions. Personalized interfaces can make those outputs feel more authoritative because they reference the user’s own portfolio.
The next behavioral risk is herding. If millions of investors rely on similar models, data sources and portfolio agents, independent decision-making can become correlated. AI may reduce the cost of research while also reducing the diversity of the research process. The practical risk is not that everyone receives exactly the same stock tip; it is that models may react similarly to the same news, rankings, momentum signals or risk constraints.
| Risk | Why AI can amplify it | Retail control |
| Automation bias | Personalized, fluent output can feel more reliable than it is | Require source checks and independent confirmation |
| Hallucination / stale data | Model can produce confident but wrong factual claims | Verify prices, filings and corporate actions at primary sources |
| Herding | Common models and data can push users toward similar conclusions | Diversify research inputs; avoid agent-only decision loops |
| Over-trading | Lower research friction can make acting feel costless | Use pre-set decision rules and cooling-off periods |
| False protection assumptions | Users may think AI output is regulated advice | Check who provides the advice and which protections actually apply |
Table 2. The biggest risks appear when users mistake convenience and personalization for regulated fiduciary judgment.
What would prove that AI has become financial utility?
The most convincing evidence will not be another survey asking whether investors ‘like AI.’ It will be observed behavior. Does AI usage remain high after a period of poor market performance? Do investors pay for AI-enhanced brokerage tiers? Does AI reduce abandonment, increase funded assets or improve retention? Do users delegate more complex tasks after first adopting low-risk research features? And, critically, do platforms publish evidence that AI-assisted investors make fewer errors rather than simply trade more often?
The autonomy ladder matters. Summarizing a portfolio is different from recommending a trade; recommending is different from automatically rebalancing; and rules-based automation is different from an agent independently deciding what to buy. Trust may increase step by step rather than all at once. The firms that respect those boundaries may build more durable adoption than firms that treat every improvement in model capability as permission to remove human control.
For the AI-stock trade, the proof points are equally concrete: revenue created per dollar of capex, cloud and inference utilization, free-cash-flow conversion, incremental margins, customer willingness to pay and whether productivity gains appear outside the technology sector. If the utility grows while those economics disappoint, AI can succeed as a technology while parts of the AI investment boom fail as investments.
The investment implication: AI is becoming a factor, not a sector
The deeper implication is that AI may be moving from an investable category to an input that cuts across categories. That transition tends to make thematic investing harder. When a technology is scarce, owning the suppliers can be the obvious trade. When the technology becomes embedded everywhere, the returns depend on bargaining power, competition, pricing and who captures the productivity gain.
Retail investors appear to be sensing that transition. They have not stopped believing in AI. In fact, Janus Henderson found 61% of affluent and high-net-worth investors expect AI to improve returns over the next five years even while 67% worry about an AI bubble in the next 12 months. Those views can coexist because one is a technology forecast and the other is a valuation forecast.
That may be the most important message behind the paradox. Mature investors do not need to choose between being ‘bullish on AI’ and ‘bearish on AI.’ They can be bullish on adoption, cautious on valuations, selective on business models and enthusiastic about using the technology themselves. That is not inconsistency. It is what happens when a market theme starts becoming an economic utility.
| Bottom Line
AI is becoming more useful to retail investors at the same time that the market is becoming less willing to value every AI-linked company on narrative alone. The widening gap between tool adoption and stock bullishness is therefore not evidence that investors have turned against AI. It is evidence that they are beginning to separate technology adoption from investment returns — a necessary transition if AI is to move from speculative theme to everyday financial infrastructure. |
Methodology and limitations
The article compares survey evidence from multiple populations. eToro’s global Retail Investor Beat surveyed 11,000 retail investors across 13 countries in August 2026; its U.S. release covers 1,000 U.S. retail investors. Janus Henderson surveyed 1,000 U.S. affluent and high-net-worth investors. Schwab surveyed nearly 1,000 retail clients, while the FCA research focused on less-experienced 18- to 40-year-old investors. These samples are not directly interchangeable, so cross-survey percentages are used to illustrate patterns rather than treated as one unified population.
The ‘utility-versus-equity gap’ is a Dave Finances calculation: the percentage using or open to AI investment tools minus the percentage expecting AI stocks to rise. It is descriptive, not a causal measure of trust. Survey wording differs between tool adoption, trust, comfort and expectations, so the article avoids treating those concepts as exact synonyms.
Hyperscaler capex estimates are forward-looking analyst estimates and can change materially. Not all hyperscaler capex is AI-related, while broader global AI investment includes spending outside the U.S. hyperscalers. These figures are used to illustrate the scale of the investment hurdle rather than to calculate a precise return requirement.
Sources and further reading
1. eToro, “Retail investors back Big Tech’s AI spending, but are more selective about which AI stocks they favour,” Sept. 30, 2026. Source
2. eToro, “Gen Z Embraces AI as Investing Tool Even as AI-Stock Enthusiasm Cools,” Sept. 30, 2026. Source
3. eToro, “Retail investors flock to AI tools, with usage up 46% in one year,” Oct. 1, 2025. Source
4. Janus Henderson, “2026 Investor Survey: Perspectives on AI,” May 19, 2026. Source
5. Charles Schwab, “Launches AI-Powered Capability That Helps Investors Understand Portfolio Performance and Market Activity,” May 5, 2026. Source
6. Robinhood, “Robinhood Cortex: The Loop Stays Dumb So the Model Can Be Smart,” Aug. 25, 2026. Source
7. Robinhood, “Puts the Power of Hedge Funds in Every Trader’s Pocket,” Sept. 29, 2026. Source
8. SoFi, “Introducing Composer by SoFi: AI-Powered Investing From Idea to Execution,” June 23, 2026. Source
9. eToro, “Tori gets real-time X intelligence, powered by Grok 4.2,” Apr. 16, 2026. Source
10. Financial Conduct Authority, “Young investors trust AI more than TV or celebrities,” Aug. 27, 2026. Source
11. Morgan Stanley, “AI Investing: Why Adoption May Drive the Next Wave,” Sept. 2026. Source
12. Goldman Sachs, “Global AI Investment Is Forecast to Exceed $1 Trillion in 2026,” Aug. 7, 2026. Source
13. Logg, Minson & Moore, “Algorithm Appreciation: People Prefer Algorithmic to Human Judgment,” 2019. Source
14. Dietvorst, Simmons & Massey, “Algorithm Aversion: People Erroneously Avoid Algorithms After Seeing Them Err,” 2015. Source
15. eToro, “US Retail Investors Call Tech Most Overvalued Sector, but Still Back AI for the Long-Term,” June 24, 2026. Source
Editorial note: This article is for informational and analytical purposes only and does not constitute investment advice. Survey responses are opinions reported at the time of collection and do not predict market returns.
Johan Shamshad is a financial markets writer at Dave Finances covering cryptocurrencies, trading platforms, brokers, fintech, financial regulation, and developments across global markets. He previously worked at Gulf News, adding newsroom experience to his coverage of fast-moving financial and digital-asset markets.
His work focuses on identifying market-moving events, company developments, regulatory changes, product launches, and shifts in trading and financial infrastructure.
Johan contributes news and analysis designed to help readers understand not only what happened, but why a development matters and how it may affect the wider financial landscape.