NextFin

Goldman Traders See Entry Point in US Stocks, but Warn of Risks

Summarized by NextFin AI
  • Goldman Sachs sees a tactical entry point after severe de-risking, but positioning data do not yet establish a durable market bottom.
  • Heavy shorting, reduced leverage, and extreme panic could fuel a rebound, although higher Treasury yields continue to pressure long-duration growth valuations.
  • The AI investment cycle is entering a structural cash-return test as slowing capital-expenditure growth and rising depreciation challenge semiconductor and infrastructure earnings.
  • The likely path is a selective, volatile recovery led by non-AI sectors, while weaker 2028 earnings expectations or higher yields could trigger another decline.

NextFin News - The US stock selloff has created the kind of tactical opening Goldman Sachs traders look for, but the same data show why calling it a durable bottom would be premature. The S&P 500 fell 3.70% on Nov. 20 to 6,910.39 in a market session that intensified pressure on technology shares, while Goldman’s trading desk put its panic measure at 9.8 out of 10. Positioning has been cut aggressively, yet not indiscriminately: short sales exceeded long purchases by 1.5 to 1, and fundamental long-short funds’ net leverage fell to 42.3%, its lowest level in a year.

The tension is straightforward. A crowded long trade can become attractive after forced selling, but a market can also fall because its earnings engine is being repriced. Goldman’s evidence points to both. The first force is cyclical and can mean-revert. The second is a structural transition in the AI investment cycle that requires a higher standard of proof.

Data cutoff: Nov. 20, 2026 market close, with Goldman and official data available through Nov. 22.

The Selloff Created an Entry Point, Not a Clean Bottom

The immediate case for a rebound is not that the macro picture suddenly improved. It is that the mechanics of the decline have become unusually one-sided. Goldman’s prime brokerage data showed a third consecutive week of global equity selling, with short sales running 1.5 times long purchases. US single-stock shorting reached its highest weekly level since June 2022, while fundamental long-short funds reduced net leverage to 42.3%. Those are the conditions in which a modestly positive catalyst can produce a disproportionate rebound because investors have already reduced exposure and bearish positions have grown.

The market proxies show how concentrated the pressure became. The SPDR S&P 500 ETF fell 3.64% on Nov. 20 to $689.81, while the Invesco Nasdaq-100 ETF dropped 5.67% to $608.45. The Dow Jones Industrial Average ETF declined 1.87% to $481.08. The gap matters: the technology-heavy proxy fell more than three times as much as the Dow proxy. NVIDIA dropped 14.01% to $125.53, a reminder that the selloff was not a broad, uniform withdrawal from equities but a repricing of the most crowded AI-linked exposures.

The official index backdrop was already fragile before the final session. The S&P 500 stood at 7,175.83 on Nov. 19, according to FRED, while the Nasdaq Composite was 23,879.20. The next day’s ETF moves show the direction and scale of the pressure without pretending that an ETF is identical to the index it tracks.

Volatility confirms stress without proving that stress has ended. The CBOE Volatility Index was 28.44 on Nov. 19, and Goldman’s desk panic measure reached 9.8 out of 10 on Nov. 20. A panic reading can become a contrarian signal when forced sellers have already acted. It can also remain elevated when investors are responding to deteriorating fundamentals rather than merely reducing exposure. The distinction is what the market must resolve next.

Goldman’s traders therefore see an entry point in selected US stocks, but they do not describe the move as a broad capitulation. The evidence of de-risking is strong; the evidence that long-term buyers have regained control is not. That makes the setup tactical before it becomes strategic.

The Transmission Mechanism Runs Through Liquidity First and Earnings Later

The first-order effect of the selloff is obvious: prices fell as investors sold AI and high-beta exposures. The second-order channel is more important. Lower leverage and heavier shorting change the market’s response to new information. If earnings, policy or company guidance merely avoid disappointment, short covering can lift prices because the marginal buyer is no longer the investor who chased the rally but the investor who must reduce a bearish position.

That is why Goldman’s entry-point argument can be right even if the market is not yet cheap on every fundamental measure. Positioning is a flow variable. It can reverse quickly. A third consecutive week of equity selling, a 1.5-to-1 short-over-long ratio and 42.3% net leverage indicate that the next dollar of demand may have more price impact than it did before the selloff. In practical terms, the market does not need a perfect earnings season to rally; it may only need fewer negative surprises than traders have positioned for.

But the flow mechanism has a limit. Short covering creates a temporary bid; it does not create revenue, cash flow or a durable return on the capital being deployed. If the next wave of information shows that AI infrastructure spending is producing lower economic returns, the same low leverage that supports a bounce can become evidence that investors have not yet finished repricing the business model.

The cross-asset confirmation is cautious rather than bullish. On Nov. 19, the 2-year Treasury yield was 4.76% and the 10-year yield was 5.04%, while the Federal Reserve’s target range stood at 3.75% to 4.00% after its Oct. 28 decision. The central bank held rates at that meeting, and two committee members preferred a 25-basis-point increase. A rebound in equities that occurs without lower financing pressure would rely on earnings and positioning, not on a wholesale easing of the discount rate. That raises the bar for long-duration growth stocks, whose valuation depends heavily on profits expected years in the future.

“The S&P 500 excluding AI is the trade,” Goldman Sachs global head of Custom Basket Trading Lou Miller said in a Goldman-published markets discussion.

Miller’s formulation captures the rotation implied by the data. The opportunity is not necessarily to restore the old leadership group; it may be to own the parts of the index whose earnings are less dependent on an uninterrupted AI capital-spending boom. The short-term trade is mean-reverting. The leadership change may not be.

AI Has Moved From a Valuation Story to a Cash-Return Test

The longer-term risk is not simply that technology shares became expensive. It is that the market must now test whether the AI investment cycle can convert extraordinary spending into durable earnings growth. Goldman Sachs Research said AI investment contributed about half of S&P 500 earnings-per-share growth in 2026. The same research expects the tailwind to fade in 2027 as capital-expenditure growth slows and depreciation expenses build.

That is a structural transition in the earnings mechanism, even if the price decline itself is cyclical. During the buildout phase, investors can reward companies for securing capacity, selling equipment and expanding infrastructure. As the installed base grows, investors ask different questions: How quickly will utilization rise? Who captures the margin? How much additional semiconductor supply arrives? What is the depreciation burden? And does spending remain economically rational if customers cannot earn adequate returns on the infrastructure they are financing?

Goldman’s market commentary shows how far parts of the trade have already adjusted. Data-center stocks were down 36% from their highs earlier in the year, memory stocks were down 42%, optical-networking shares were down 46% and NeoCloud companies were down 57%. Those losses lower the bar for positive news, but they do not answer whether the cash-flow forecasts embedded in the remaining valuations are realistic.

The second-order market effect reaches beyond semiconductors. If investors begin to doubt the pace of AI capital expenditure, the first hit lands in hardware and infrastructure. The next hit can land in industrial suppliers, utilities and credit markets that financed the buildout. A third effect can run through the dollar: Goldman analysts said US tech exceptionalism and equity inflows have supported the currency, so a weaker AI investment impulse could remove part of that support. The market could therefore experience an equity rotation and a currency adjustment at the same time.

Goldman’s 2027 outlook remains constructive in aggregate. Its research team projects an 18% total return for the S&P 500 through the end of 2027, as of Nov. 17, while expecting earnings rather than expanding valuation multiples to drive the gain. That forecast is not a contradiction of the trading desk’s caution. It says the index can rise even as the leadership model changes, provided earnings growth survives and multiples do not need to expand.

The key expectation gap is that a positive index forecast does not require every high-growth group to recover. A rotation into the S&P 500 outside AI, healthcare or other defensive growth sectors could produce better market breadth without restoring the most speculative parts of the infrastructure trade. That is a different bull case from the one investors followed earlier in the year.

“AI investment has driven a surge in S&P 500 earnings in 2026, but that tailwind should begin to fade in 2027,” Goldman Sachs chief US equity strategist Ben Snider wrote in the firm’s research.

The sentence separates the cyclical and structural stories. A short-covering rally can begin while the AI earnings impulse is fading. The rally would then be real, but its composition would matter more than its headline index return.

The Strongest Bear Case Is That Deleveraging Is Only Half Finished

The strongest counter-thesis is not that stocks can never rebound after a selloff. It is that the selloff reflects a rational deterioration in forward earnings, and the apparent panic is not an exhaustion signal but an early-stage reset. If AI capex growth slows while depreciation and semiconductor supply rise, profit estimates can fall faster than prices have adjusted. In that case, lower leverage does not mark the end of forced selling; it marks the first wave.

This case has institutional support inside Goldman’s own outlook. Snider identified a deterioration in the 2028 earnings outlook as the biggest downside risk to equities in 2027. The warning is consequential. Markets discount future profits, so a strong 2027 earnings year could fail to protect stocks if investors conclude that the following year brings slower demand, excess capacity or margin compression. A market can look through a weak quarter; it struggles to look through a broken multi-year earnings path.

The bear case also explains why elevated positioning may not be fully cleared. The prime brokerage figures show aggressive selling, but broader capitulation may still be missing. Short sales outpaced long purchases, but the existence of many shorts can support a rally if the news improves. It can also amplify another leg lower if the market receives a new negative catalyst and shorts add to positions rather than cover them.

The falsifying signal for the tactical-bottom thesis is quantifiable: if the S&P 500 closes below 6,910.39 while the VIX remains above 30 and Goldman’s net-leverage measure falls below 42.3% without a corresponding improvement in earnings revisions, the evidence would favor unfinished fundamental deleveraging over a completed panic. The opposite signal would be a rebound accompanied by improving forward earnings revisions and a decline in volatility, not merely a one-day short-covering surge.

The counter-thesis should not be dismissed because the market has already punished AI-linked shares. A 57% decline in NeoCloud stocks or a 46% fall in optical networking can make a stock less crowded while leaving its end-market assumptions wrong. Price damage changes the entry valuation; it does not repair a demand forecast. The bear case wins if the market discovers that the return on new capital is structurally lower than investors assumed.

Still, the bear case has a weakness. It treats the entire US equity market as a single AI duration trade, while Goldman’s rotation work points to a broader opportunity outside AI. Miller described companies with high-teen multiples and low-teens earnings growth as a way to own quality compounders in a slowing-growth environment. The index can stabilize even while the most exposed infrastructure names continue to struggle. That does not make the market immune to a recession or a rates shock. It means the distribution of risk has changed.

What the Entry Point Means Across Time Horizons

In the short term, the dominant variable is liquidity. The 3.64% fall in the S&P 500 ETF proxy and 5.67% decline in the Nasdaq-100 ETF proxy, combined with a 9.8 out of 10 Goldman panic reading and 42.3% net leverage, create the ingredients for a relief rally. The base case is a volatile rebound led by heavily shorted and recently de-rated sectors, but with leadership rotating rather than returning wholesale to the prior AI winners.

The upside scenario requires a specific catalyst: company results or guidance must show that AI demand remains strong enough to sustain investment while margins hold. If that happens and the VIX falls below 25, short covering could broaden from infrastructure into quality growth and cyclical exposures. The move would be strongest where positioning is most defensive, not necessarily where the long-term fundamentals are best.

The downside scenario requires no spectacular collapse. A renewed rise in long-term yields, weaker 2028 earnings expectations or evidence of excess semiconductor supply could push investors to sell the rebound. The 10-year Treasury yield at 5.04% on Nov. 19 shows that equity investors are not operating against a falling-rate backdrop. If yields move materially higher while earnings revisions weaken, the discount-rate and earnings channels would reinforce each other.

In the medium term, the beneficiaries are companies whose profits depend on broad economic demand, disciplined capital allocation and less aggressive AI infrastructure assumptions. Healthcare is one candidate Goldman has highlighted because it combines defensive characteristics with potential technology-driven growth. The exposed group includes data-center developers, high-beta memory and optical suppliers, and companies whose valuations require uninterrupted capital spending by a small group of hyperscale customers.

In the long term, the question is structural: does AI become a productivity platform with returns that justify the buildout, or a capital cycle that produces excess capacity before demand catches up? Goldman’s global outlook forecasts 2.8% growth in 2027, only modestly above 2.7% in 2026. That steady macro path is supportive but not powerful enough to absorb every valuation mistake. The market impact of AI can be much larger than its direct GDP contribution because it is concentrated in capex, earnings and market value.

Investors will therefore watch three linked signals rather than one index level: forward earnings revisions for 2028, evidence of utilization and pricing in AI infrastructure, and the behavior of long-term Treasury yields. A falling VIX by itself would show that fear has eased. It would not show that the earnings transition has been solved.

The Goldman traders’ message is best understood as a timing judgment, not an all-clear. The selloff has made selected US stocks more interesting because positioning has become defensive and short exposure has risen. But the market still needs proof that the AI investment cycle is moving from spending to cash generation.

That leaves three scenarios. The base case is a choppy, selective recovery as short covering meets slower earnings growth. The upside case is a breadth-led rally if non-AI earnings stabilize and AI demand remains profitable. The downside case is another leg lower if 2028 estimates deteriorate, long-term yields rise and the next wave of selling comes from fundamental investors rather than leveraged funds.

Goldman’s entry point is real, but it belongs to a market between two regimes: the panic may be cyclical, while the earnings test is structural. The rebound can start before the repricing is finished.

Explore more exclusive insights at nextfin.ai.

Insights

What triggered the US stock selloff?

Why does Goldman see a tactical entry?

How does short covering lift prices?

What does 42.3% net leverage signal?

Why are AI stocks hit harder?

How do Treasury yields affect stocks?

Is the selloff cyclical or structural?

Why may AI spending growth fade?

Can AI capex produce durable returns?

Could signals falsify the bottom thesis?

Which sectors may lead the rebound?

Could AI weakness hurt the dollar?

Which signals confirm durable recovery?

How does AI compare with past cycles?

What did the Fed decide on Oct. 28?

Could excess supply trigger another drop?

Can index gains hide leadership change?

What is Goldman 2027 S&P forecast?

Search, ask or analyze anything...