Goldman Sachs Warns Momentum Divergence Hits Five-Year High as AI Stocks Suffer Record Drawdown
Goldman Sachs strategist Guillaume Soria warned in a client note on September 15 that AI-related trades are facing an unprecedented internal split in momentum factors. The 3-month momentum index rose 5% on the day, while the 12-month momentum index fell 6.7%, marking the widest single-day performance gap in five years. Meanwhile, Goldman's index tracking AI-related stocks has dropped nearly 45% from its high, the largest drawdown since the debut of ChatGPT.
The note suggested that investors with AI exposure could hedge by buying put options on the medium-term winners basket (GSXUHMOM) or the AI beneficiary basket (GSTMTAIP), citing the high overlap between AI trades and momentum strategies. Goldman noted that one-year positioning in the momentum factor has fallen to the 41st percentile, while the five-year reading remains elevated at the 88th percentile, suggesting systemic deleveraging has yet to fully play out. Momentum factor volatility was also notably higher than the broader market, with the S&P 500 down less than 30 basis points on the day.
Significant rotation is underway within momentum portfolios. Goldman data showed short-term momentum capital flowing from semiconductors to software, a sector that was among the most unloved at the start of the year. The bank's software-vs-semiconductors relative performance index posted its second-largest single-day gain on record. Software has long been the largest short position in the 12-month momentum basket; its rapid rebound may lead to future momentum basket rebalancing, with buying of software and selling of semiconductors.
Goldman said that although the correlation between AI stocks and momentum factors remains high at 90%-96% over the one-month to one-year horizon, this long-standing relationship is showing signs of loosening. If the rotation from semiconductors to software continues, the industry composition of momentum baskets will keep shifting, prompting investors to reassess strategies that use AI as a proxy for momentum exposure.
