AI Boosts Market Efficiency but Heightens 'Tail Risk,' Bernstein Says

Artificial intelligence is making financial markets more efficient in many respects, but it is also introducing new vulnerabilities that could amplify volatility and sharp market downturns, according to a research note from Bernstein.

The Efficiency Gains

Analysts at Bernstein noted that at the company level, AI is improving market efficiency by processing vast amounts of data—including earnings reports, regulatory filings, and alternative datasets—faster than humanly possible. This acceleration helps to narrow information gaps, improve price discovery, and reduce the magnitude of earnings surprises as analyst forecasts converge more quickly with actual results.

Beyond pure speed, AI is also transforming the research process. Bernstein highlighted how automated workflows are cutting the time analysts need for routine tasks like earnings reviews and financial model updates. This allows for expanded coverage, with the note pointing to a sharp rise in analyst coverage of emerging market small-cap stocks over the past year, a trend that could eventually reduce the valuation premium often associated with under-researched companies.

New Risks and Vulnerabilities

Despite the benefits, the note warns that the widespread adoption of similar AI models creates new systemic risks. As more investors use comparable tools trained on overlapping data, trading signals can become increasingly synchronized, leading to crowded positions and making market reversals more severe during periods of stress.

Ad

Bernstein pointed to recent events as potential examples of these risks, citing the August 2024 unwind of the yen carry trade and instances of AI-generated misinformation that briefly moved U.S. equities. These examples illustrate how automated strategies and synthetic content can accelerate volatility before human investors have time to verify new information.

Another concern raised is the "reflexivity problem," a feedback loop where AI-generated recommendations influence investor behavior, pushing prices in a direction that then reinforces the models' future outputs. According to the note, this could strengthen momentum trades, increase market concentration in a few names, and produce larger valuation extremes.

A Market of Extremes

The overall conclusion from Bernstein suggests a new market paradigm is emerging. AI is likely to reduce average, day-to-day market inefficiencies through superior research and faster execution. However, this comes at the cost of increasing the likelihood of larger, more severe dislocations during periods of market stress, resulting in lower average inefficiency but significantly greater tail risk.