This month it has become clear that something big is happening underneath the market. AI is causing uncertainty on the value of long-duration software assets, forcing a re-rating as investors question the disruption of AI across sectors. At the same time, a surprise rise in PMIs has helped accelerate a rotation from growth to value, creating cross-currents that are destabilizing relationships across sectors, factors, and stocks. Early in February I got a warning signal on my turbulence model and posted it on X. The warning requires three conditions to fire simultaneously: first, a covariance shock, which simply means assets that normally move independently are suddenly moving together in unusual and volatile ways, breaking the “normal” relationships that risk models depend on; second, VIX remaining low, indicating the options market isn’t yet pricing in fear; and third, SPX still trading above its 50-day moving average, meaning price action still looks healthy. When all three conditions occur together, it signals hidden stress that hasn’t yet surfaced in traditional indicators, the kind of divergence that historically precedes larger dislocations.
The numbers are becoming alarming. From 2023 through 2025 – a span of roughly 28 months – the model generated approximately 20-25 warning signals, or less than one per month. In the first six weeks of 2026 alone, I’ve recorded 12-15 warning signals – a rate of 8-10 per month, with most of those occurring this month, representing a 10x increase in frequency. The market keeps experiencing stress, appearing to recover, then experiencing stress again, with each cycle potentially exhausting its resilience.
The low covariance environment of the last couple of years allowed gross leverage to rise to very high levels, as evidenced by prime brokerage data showing near-record positioning. When correlations are stable, risk models give the green light to add exposure – and funds did exactly that. Normally, to see a covariance shock, there is a monetary (rate hikes), fiscal (tariffs), or credit event. Usually, these are temporary. This one, in my opinion, as I have written, is different. This all fits with Elon Musk’s description of AI as a “supersonic tsunami.” AI speed, deflationary pressure and its reality is still doubted by most and positioning around the globe is still in massively in growth and companies built on code which is now free and ubiquitous.
A force moving this fast (supersonic) and this powerful (tsunami) doesn’t just disrupt one sector; it reshapes the entire landscape of winners and losers simultaneously, creating exactly the kind of cross-asset correlation shocks that destabilize risk models. As of now, it has been isolated to stocks, but given the sustained covariance shocks, it is always important to see if there is deleveraging contagion, and there are now small signs. Credit has slowly been bleeding over the last couple of weeks, adding another crack in the foundation. With constant turbulence signals, a structural force like AI reordering the economy and creating uncertainty on long-duration assets, leverage at elevated levels, and crowded positioning across the industry, I expect the market to remain volatile all year. Be nimble and get ready for a year where every month feels like a year.
Source: 22V Research
What is the Turbulence Model
The turbulence model itself is built to detect instability in the covariance matrix that underpins modern portfolio risk management. Hedge funds depend on stable correlations between assets to calculate position sizes, determine leverage ratios, and ensure that diversification actually provides protection. When these correlations are stable, leverage appears “safe” and hedges work as intended. The turbulence score measures when this stability breaks down, when assets that normally move independently start moving together in unusual ways, invalidating the assumptions baked into risk models and leading to increased portfolio volatility. The green bars on the chart represent the most dangerous phase: the period when the covariance matrix is destabilizing but market prices haven’t reflected it yet (SPX still rising, VIX still low). This is the window before forced deleveraging begins – when risk models are flashing internal warnings but the selling cascade hasn’t started. The fact that 2026 has produced as many green bars in six weeks as the prior two years combined suggests the covariance matrix is repeatedly destabilizing, risk models across the industry are getting hammered with warnings, and each “recovery” is burning through the system’s resilience, exactly the pattern that precedes correlation-driven meltdowns.
