QRAFT AI-Enhanced U.S. Large Cap ETF (QRFT)

The QRAFT AI-Enhanced U.S. Large Cap ETF (QRFT) applies machine learning algorithms to the task of picking and weighting large-cap U.S. stocks. Rather than mirroring a fixed index by market capitalization or equal weighting, the fund uses data patterns and predictive models to identify which large companies are most likely to outperform, then positions the portfolio accordingly. The result sits between a passive index fund and an actively managed mutual fund—using systematic, automated decision-making rather than a human stock picker’s judgment, but aiming to beat a plain benchmark through AI-driven stock selection.

QRAFT, a Seoul-based investment technology firm, developed the proprietary machine-learning framework behind the fund. The approach ingests historical data on stock fundamentals, valuations, earnings trends, price momentum, and other quantifiable factors, then runs these through neural networks and ensemble models to forecast which stocks are most likely to outperform. The algorithm is rebalanced regularly (typically monthly or quarterly), allowing the weightings to shift as the model’s predictions change. QRFT holds typically 80 to 120 large-cap U.S. stocks drawn from the S&P 500 and Russell 1000, weighted according to how bullish the AI model is on each one.

The appeal is intuitive. If machine learning can beat humans at predicting chess moves or recognizing faces, why not stock performance? Large-cap stocks generate mountains of public data—quarterly earnings, analyst forecasts, insider trading, price action—and the patterns hidden in that data could theoretically guide a better-than-random stock selection process. Where a human investor might overlook a subtle shift in a company’s competitive position, a trained model might spot it in the numbers. The fund sits in a growing category of “smart beta” and “AI-enhanced” ETFs, each claiming that computers can see what traditional indices miss.

The mechanics matter more than the promise. QRFT carries an annual expense ratio in the 0.65 to 0.75 percent range—higher than a passive S&P 500 tracker (which costs around 0.03 percent) but cheaper than a traditional actively managed large-cap fund (often 0.75 percent or more). The fund is transparent about its holdings, disclosing them regularly so you can see exactly which 80-120 stocks it holds and how heavily it is positioned in each. Trading liquidity is solid; the fund has enough daily volume that bid-ask spreads are tight.

The critical question is whether the AI model actually outperforms. In its early years, QRFT showed promise—delivering returns competitive with the S&P 500 with somewhat lower volatility in some periods. But the fund’s track record is still short (it launched in 2021), and it has not yet lived through a full market cycle, multiple regime changes, or an extended period where the model’s predictions prove systematically wrong. Machine-learning models trained on historical data carry the risk of overfitting—appearing to work brilliantly on the data they were trained on but failing in new conditions the training data never saw. A bull market that plays to different factors than the model learned, or a shift in which stocks drive the broader market, could flip QRFT’s edge into a drag.

There is also the concentration risk embedded in any AI selection process: if the model converges on similar trading ideas as other momentum or value factors, QRFT could end up heavily tilted toward a handful of favored themes or companies, amplifying drawdowns in down markets. And unlike a passive index, which you understand by construction (it holds all 500 S&P stocks, weighted by size), QRFT’s exact logic is opaque—the fund explains that it uses machine learning but not how the model makes its decisions at a granular level. Investors must trust both the algorithm and the firm’s ability to maintain and adapt it over time.

For researchers, the fund’s fact sheet and prospectus lay out the basic strategy and holdings. The real work lies in comparing QRFT’s returns (gross and net of fees) to relevant benchmarks—the S&P 500, the Russell 1000, or other large-cap indices—over as many years of history as the fund has available. Watch whether QRFT tends to outperform in growth years and underperform in value years, or vice versa, which would signal what factor tilts or sector bets the model favors. Also examine the fund’s turnover (how often it trades in and out of stocks); high turnover can create tax drag for taxable accounts. The promise of AI-enhanced investing is real, and the approach is systematic and scalable—but QRFT is still being tested by live markets, and investors should treat it as an experiment with genuine upside potential and genuine execution risk.