AI-Driven Investing: Which Small-Cap Funds Outperformed the Russell 2000?

According to Research and Markets, the market for artificial intelligence systems used in asset management grew from $5.39 billion in 2025 to $7.1 billion in 2026, and by 2030, it could reach $21.82 billion, representing a compound annual growth rate of 32.4%.

In 2025, 95% of the fund managers surveyed used generative AI in their work, compared to 86% in 2023, according to data from the industry association AIMA. Moreover, there is demand from clients as well: 60% of institutional investors stated that they are more likely to invest in a fund that allocates a significant budget to research and the implementation of generative AI.

The first exchange-traded fund with a portfolio built using an AI model debuted on the exchange in October 2017—the Amplify AI Powered Equity ETF (AIEQ). Developed by EquBot, the fund runs on the IBM Watson platform and processes data from approximately 6,000 U.S. companies daily, including financial reports, news, analytical materials, and social media posts. Since its inception, the fund has generated a total return of about 130%.

Over the past nearly ten years, the selection of AI-managed funds has expanded significantly. For example, WisdomTree and South Korea’s QRAFT offer such funds. VanEck offers the BUZZ index fund, which uses analysis of investor sentiment on the internet. In February 2026, Pictet launched the PQUS fund, in which securities are selected by a machine-learning model. The model was trained on approximately 400 company characteristics drawn from 15 years of market history and is retrained every quarter. However, the model does not make all decisions on its own: portfolio managers set the portfolio constraints themselves—for example, by factors, sectors, and countries. The model is responsible only for selecting specific securities within these parameters.

To assess how successfully AI performs as an investor, Oninvest selected 19 investment tools that use AI and machine learning to select stocks, forecast returns, analyze financial and text data, and determine the weightings of securities in a portfolio.

They can be broadly divided into two groups: 11 small- and mid-cap funds and 8 broad-market strategies that are not limited by company size. In the first group, there is only one ETF; the rest are mutual funds. In the second group, all are ETFs.

From the beginning of 2026 through August 4, the BlackRock Advantage SMID Cap Fund posted the best performance in the sample, gaining 25.3%. It was followed by the BlackRock Advantage Small Cap Core Fund, which rose 24.3%. The Russell 2000 Index rose 22.36% over the same period. Only these two funds managed to outperform the benchmark. The remaining strategies underperformed the index, including the Federated Hermes MDT Small Cap Value Fund, the Voya MI Dynamic SMID Cap Fund, the Federated Hermes MDT Small Cap Core ETF, and the Federated Hermes MDT Mid Cap Growth Fund.

By comparison, the picture is the opposite among AI strategies from large companies: five out of eight funds outperformed the S&P 500 Index, which posted a return of 13.02%. The best results were posted by the QRAFT AI-Enhanced U.S. Large Cap Momentum ETF, which rose 21.29%, and the WisdomTree International AI Enhanced Value Fund, which rose 19.47% (fund returns are calculated on a reinvested dividend basis).

We selected the most interesting funds: one from each major provider in the small- and mid-cap segments and, for comparison, one exchange-traded fund from the broad-market group.

Year-to-date growth: 20.8%

The only full-fledged ETF in our SMID group. Federated Hermes launched it in July 2024 alongside three other quantitative strategies, bringing the MDT approach—which the company has been developing for over 30 years—to the exchange.

The model ranks stocks using several algorithms that evaluate companies based on various parameters. Morningstar analyst Drew Carter, in his analysis of a mutual fund built on the same MDT approach, rated the investment process as above average. In his view, the strategy combines value and quality signals with technical indicators into a single, flexible machine-learning system. However, he rated the performance of the management team as only average.

As of July 31, 2026, the ETF’s assets totaled $309.4 million, and its net expense ratio of 0.36% is the lowest in our group. The portfolio is well diversified and has no pronounced sector bias. Among the largest holdings are American Healthcare REIT, which owns nursing homes and medical centers; Bloom Energy, a manufacturer of fuel cells for data centers; BrightSpring Health Services, a provider of home care services; and Powell Industries, a manufacturer of electrical distribution equipment.

Year-to-date growth: 25.3%

The top performer in the sample and the only fund to significantly outperform its segment: since the beginning of the year, it has returned 25.3%, compared to 22.36% for the Russell 2000. Formally, this is an institutional-class mutual fund with a management fee of 0.48%; its benchmark is the Russell 2500.

In the prospectus filed with the U.S. Securities and Exchange Commission, BlackRock explicitly states that it uses, among other things, machine learning and AI—including large language models and text sentiment analysis—to forecast returns. The asset management firm describes the strategy as a low-cost portfolio with a technology-driven approach to stock selection.

At the same time, the allocation to individual securities is small: no single position accounts for more than 1% of assets. Among the largest holdings are the Texas-based Cullen/Frost Bankers, the HVAC contractor Comfort Systems USA, the aircraft bearing manufacturer RBC Bearings, and Carpenter Technology, a producer of specialty alloys.

Year-to-date growth: 19.47%

From the broad-market group, we singled out the WisdomTree International ETF—the only fund in our sample that invests outside the U.S. In 2025, it posted the best return among all 19 funds analyzed by Oninvest—38.69%. Since the beginning of 2026, it has gained 19.47%. Within its group, the fund was outperformed only by the QRAFT AI-Enhanced U.S. Large Cap Momentum ETF, which rose 21.29%.

In January 2022, WisdomTree transitioned two of its dividend funds to strategies based on Equity Machine Intelligence—a model used by the Voya MI Dynamic SMID Cap Fund. WisdomTree International AI uses it to identify undervalued stocks in developed markets outside the U.S. WisdomTree describes the model as an “analyst” with access to 20 years of market data and approximately 10,000 metrics for each company.

The portfolio is dominated by large European and Asian companies: Singapore’s United Overseas Bank, France’s TotalEnergies, Sweden’s Swedbank, British American Tobacco, and National Australia Bank; the fee is 0.58%.

The U.S. counterpart to WisdomTree International AI rose by only 9.72% in 2025. This illustrates that a strategy’s performance is determined not only by the model itself, but also by the market in which it operates.

The strong performance of AI-managed small- and mid-cap funds in 2026 is primarily due to the growth of the segment itself, rather than superior stock selection. Only two out of nine funds outperformed the Russell 2000, whereas in the group of strategies focused on large-cap companies, five out of eight outperformed the S&P 500.

At the same time, the potential for successfully selecting small-cap stocks is indeed higher. As far back as December 2025, Goldman Sachs noted that the range of returns within the Russell 2000 is more than double that of the S&P 500. This means more opportunities for portfolio managers to find stocks capable of generating excess returns relative to the index.

It is also important to understand what these products are not. There are virtually no fully autonomous funds: according to a 2026 Mercer survey, only about 5% of asset managers grant their models autonomous or partially autonomous authority to make investment decisions.

At WisdomTree, the selection process involves five stages of human review, while at BlackRock, an AI model is integrated into the portfolio optimization process but does not replace it. In its third-quarter 2025 letter, the Voya team explicitly acknowledged that the fund’s underperformance relative to the Russell 2000 was due specifically to its selection of individual stocks.

These funds should be viewed primarily as active quantitative strategies, rather than as a bet on the technology itself. In its January 2026 outlook, Two Sigma noted that success depends just as much on the quality of research and the team’s capabilities as it does on the level of technology used.

For investors, the rule is simple: first, look at the market, the fees, and the portfolio’s composition; only then should you consider the model used to select stocks. The average fee for equity funds in the U.S. is 0.4% per year, while that for index exchange-traded funds is 0.14%. Against this backdrop, paying 0.7–0.86% per year—as is the case with certain funds from Voya and Federated Hermes—for returns that fall short of the benchmark is hardly justified.

This is not intended as individual investment advice.