Goldman Flows Guru Warns Relatively Calm Index Belies 'Fragile' Market With 'Poor Tolerance For Bad News'

The market reaction to the Middle East headlines so far has been fairly consistent with how equities historically trade around geopolitical shocks.

Since 1950, during seven major spikes in the Geopolitical Risk Index, the S&P 500 has fallen roughly 4% on average in the first week, but has typically recovered back to pre-shock levels within the following month.

The distribution of outcomes is wide, but the key point is that markets typically price the uncertainty quickly before refocusing on growth and earnings.

But, while the S&P’s max drawdown was still only -3.4%, Goldman Sachs US Vol Panic Index closed the week at 9.72 on Friday – a pretty notable gap between the level of internal stress being priced and the damage actually showing up at the index level.

As Goldman flows guru, Lee Coppersmith puts it, the panic signal is behaving more like a market under much heavier pressure than the headline tape would suggest.

That keeps the message the same as it’s been in prior episodes: beneath a relatively calm index backdrop, the market is still trading with elevated fragility, poor tolerance for bad news, and a much fatter left tail than SPX alone implies. Until that gap closes, I think it argues for respecting downside convexity even if the index itself hasn’t fully cracked yet.

Positioning and flow data reinforce the idea that investors are hedging more, but not meaningfully backing down. US equities were net sold for a third straight week (-0.2 SD vs. the past year), driven primarily by selling in macro products, while single stocks were actually net bought for the first time in five weeks (+1.2 SD). That dynamic was most evident in ETFs, where US-listed ETF shorts jumped +8.3% on the week – the largest increase since the week of Liberation Day and the second largest in the past five years – pointing to a meaningful pickup in hedging activity across Corporate Bond, Energy, Small Cap Equity, and Large Cap Equity ETFs.

At the same time, gross leverage only eased modestly to 307.4% (-0.7 pts, still in the 94th percentile vs. the past year / 99th percentile vs. five years) while net leverage fell just -1.0 pt to 79.2%, which suggests investors are adding protection around the edges rather than materially de-risking core exposure.

Under the surface, we also saw some early signs of positioning adjustment. While US TMT stocks were net bought this week, the activity was driven almost entirely by short covers, pointing more to risk unwinds than true conviction. In fact, the week’s de-grossing in US TMT was the largest since July ’24 and one of the biggest in the past five years. Exposure still remains elevated following 22 consecutive weeks of rising gross trading flows, with aggregate TMT exposure still sitting in the 97th percentile versus the past year. At the same time, managers continued rotating away from cyclicals, which were net sold for a second consecutive week, leaving hedge funds modestly underweight the group relative to the Russell 3000.

PB positioning also shows how crowded leadership has become. Net exposure to the Medium-Term Momentum factor across GS PB books is now pushing ~55–60% of fund equity – essentially the highest reading in years and near the top of the historical range.

In other words, the market is still heavily concentrated in the winners that have driven this cycle. That kind of crowding rarely matters on the way up, but it does tend to matter if/when leadership starts to crack.

The crowding is already starting to show up in hedge fund performance. The most common feedback we heard this week was simply how hard the market was to navigate. Our GS Fundamental L/S Performance Estimate fell -3.22% between Feb 27 and Mar 5 (vs MSCI World -2.33%) – the worst weekly drawdown since June ’22 – with both beta and alpha contributing, and alpha posting its weakest week since Jan ’22 as both longs and shorts struggled. Even systematic L/S books lost ground (-0.45%), reinforcing just how difficult the tape has become.

Looking under the surface, the winners vs. losers dynamic was even more violent than the index suggests. Some of the market’s biggest YTD winners – Global Copper (-14.7%), Memory (-13.6%), and US Metals (-10.2%) – were among the week’s worst performers, while some of the most beaten-up areas of the market caught a bid, including Software-at-Risk (+7.4%), Expensive Software (+6.9%), and AI-at-Risk (+5.5%). It had the feel of a classic factor unwind – crowded winners getting hit while laggards rallied – highlighting just how sharp the rotation was beneath a relatively contained index move.

One thing that surprised me this week was how resilient EM flows remain. GEM funds have now seen 20 straight weeks of inflows, with $57bn added YTD – the fastest pace of buying in two decades. Despite the volatility earlier in the week, our prime desk also didn’t see meaningful de-grossing in Asia, with flows broadly flat and positioning still elevated. PB allocations to Korea and Taiwan remain near recent highs, with Korea net exposure still around ~5% of fund equity across the global prime book.

That said, our research team is pushing back a bit on the more bearish Korea narrative. The recent ~20% pullback in KOSPI needs to be viewed in the context of the +176% rally since April ’25, and they see the move more as a correction than the start of a bear market. Positioning also looks less stretched than many assume – foreign investors have actually net sold ~$15bn YTD, retail leverage remains modest (~0.6% of market cap), and domestic institutions have been steady buyers. Meanwhile fundamentals continue to improve: the team raised its 2026 earnings forecast to +130% on strong memory pricing and keeps an OW rating with a 7000 KOSPI target (~25% upside).

Finally, it’s worth remembering just how concentrated the market’s earnings engine has become.

Seven AI-exposed stocks (AMZN, AVGO, GOOGL, META, MSFT, MU, NVDA) accounted for roughly half of S&P 500 EPS growth in 2025, and are expected to drive a similarly large share again in 2026. NVDA alone is projected to account for roughly 24% of total S&P EPS growth next year.

In other words, even as macro headlines dominate the tape, the earnings engine of the index remains overwhelmingly tied to the AI capex cycle.

Ultimately, what ties all of this together is the gap between macro resilience and market fragility. Historically geopolitical shocks tend to fade and equities refocus on growth and earnings fairly quickly. But the current setup is complicated by positioning and concentration. Investors are still running high leverage, leadership remains extremely crowded, and the earnings engine of the index is increasingly dependent on a very small group of AI-linked companies. That combination helps explain why the index drawdown has been relatively contained while the internals have felt far more unstable.

Put differently – this shock may ultimately fade like others before it. But until positioning de-crowds or leadership re-accelerates, the combination of crowded factor exposure, elevated leverage, and extreme earnings concentration likely keeps the tape choppier than the index alone would suggest.

Professional subscribers can read much more from Goldman's Sales & Trading team here at our new Marketdesk.ai portal