Is the AI trade over?

Artificial intelligence has been the defining equity-market theme of the past several years. The companies viewed as the clearest beneficiaries of AI attracted enormous amounts of capital, while businesses thought to be vulnerable to AI disruption were often treated very differently.

That distinction produced one of the more powerful relative trades in the market.

Goldman Sachs tracks it through a pair that is effectively long its AI Beneficiaries basket and short its AI At-Risk basket. The former includes companies expected to benefit from or enable AI; the latter includes companies whose businesses investors believe could be disrupted by it.

For several years, this trade was easy. But a funny thing happened on the way to the checkout. By the end of the second quarter of this year, the AI beneficiaries started sputtering—that’s not a shock; sometimes trades simply need to take a breather. The more surprising thing is that the AI At-Risk companies suddenly rose from the dead, starting roughly around the beginning of the third quarter.

We didn’t use the Goldman indices to make our own assessment, but we did borrow from their themes.

We constructed two equal-weighted, 20-stock baskets designed to approximate the distinction between AI Beneficiaries and AI At-Risk companies. The Beneficiaries basket consists of companies positioned to capture incremental demand from AI adoption (e.g., companies providing semiconductors, computing infrastructure, cloud platforms, networking, power, and other enabling technologies). The At-Risk basket, on the other hand, focuses on companies whose existing business models may face greater disruption from AI, particularly where revenues depend heavily on human labor, traditional information-processing activities, or services that generative AI can increasingly automate or disintermediate.

The baskets are intended as transparent proxies for the investment themes, not replications of Goldman Sachs’ proprietary indices, with equal weighting used to prevent a handful of mega-cap companies from dominating the comparison.

Something has clearly changed. This, however, doesn’t mean the AI story is over. Anything but. It does mean we should probably distinguish between the long-term economic significance of artificial intelligence and the much narrower investment proposition that the same group of stocks will continue outperforming the companies presumed to be threatened by it.

The relative trade has reversed sharply.

Chart 1

AI Beneficiaries Have Dominated—But the Gap Has Recently Begun to Narrow

Cumulative returns of equal-weighted AI Beneficiary and AI At-Risk baskets, April 30, 2024–September 15, 2026

From April 30, 2024 through September 15, 2026, our equal-weighted basket of AI Beneficiaries gained 191.6%, while the equal-weighted AI At-Risk basket declined 24.8%. Both series are shown on a cumulative-return basis from the same starting point. Source: LucenWorks; Belvest Family Wealth basket construction.

The first chart provides a more intuitive way to see what has happened than a relative-performance index alone. We constructed two equal-weighted baskets: companies whose businesses appear positioned to benefit directly from the development and adoption of artificial intelligence, and companies whose existing economics may be more exposed to AI-driven disruption.

The divergence has been extraordinary. Since April 2024, the Beneficiaries basket has nearly tripled, while the At-Risk basket has lost roughly one-quarter of its value. In other words, the AI trade was not simply about owning companies participating in an important new technology. Investors were simultaneously rewarding the perceived winners and penalizing the perceived losers.

For most of the period, that judgment worked remarkably well. But the right-hand side of the chart is more interesting. The Beneficiaries remain dramatically ahead over the full period, but the At-Risk companies have rebounded from their lows at the same time that the Beneficiaries have become considerably more volatile. The gap has begun to close.

To isolate that change, we can take the analysis one step further.

Chart 2

The AI Winners-versus-Losers Trade Has Reversed Sharply

100% long AI Beneficiaries / 100% short AI At-Risk; 0% net exposure and 200% gross exposure, July–September 2026

A portfolio constructed with 100% long exposure to our AI Beneficiaries basket and 100% short exposure to our AI At-Risk basket lost 31.1% over the period shown through September 15, with a maximum drawdown of 39.4%. Source: LucenWorks; Belvest Family Wealth basket construction.

This second exhibit requires some of the same abstract reasoning as international investing when we introduce a currency numeraire: instead of asking how either group performed in isolation, we are asking how one performed in terms of the other.

The portfolio begins with no net equity exposure—100% long Beneficiaries and 100% short At-Risk—but 200% gross exposure. A decline therefore means that the relationship between the two groups is moving against the trade: the Beneficiaries are underperforming the At-Risk companies, whether because the former are falling, the latter are rising, or both. Recently, that reversal has been severe.

The long-Beneficiaries/short-At-Risk portfolio fell 31.1% over the period shown and, at its low, was down 39.4%. It subsequently recovered some of that loss, but the magnitude of the move is difficult to dismiss as ordinary noise around an otherwise unchanged trend.

We don’t think that says artificial intelligence has somehow become less important, nor does it mean the companies in the Beneficiaries basket have suddenly become poor businesses. The first chart makes clear just how extraordinary their prior gains have been.

Instead, the two charts together suggest that the market may be reconsidering a much more specific proposition: that the companies expected to benefit from AI should continue to outperform, almost mechanically, the companies expected to be disrupted by it. After such an enormous divergence, that distinction matters. Valuations may have moved too far, expectations for disruption may have become too severe, or some of the supposedly vulnerable businesses may be proving more adaptable than investors initially assumed. The recent reversal could also simply represent an unusually violent correction within a longer-term trend.

We don’t know which interpretation will prove correct, but what we do know is that a relationship that had been extraordinarily powerful and persistent has changed direction sharply enough to merit attention. The long-term AI story is still being told, even if most of the Stark family seems to be on the run.

Sentiment has suddenly fallen through the floor; investors are nervous.

Chart 3

Individual Investors Have Turned Sharply Bearish

AAII Investor Sentiment Survey, bullish versus bearish respondents, 2022–2026

In the latest survey, 53.3% of respondents were bearish and 28.8% bullish, producing a bull-bear spread of -24.5 percentage points. Source: American Association of Individual Investors.

This is where the story becomes a little more complicated. AAII’s weekly survey has suddenly become quite pessimistic. More than half of respondents expect stocks to decline over the next six months, while fewer than three in ten describe themselves as bullish.

Sentiment measures are often used as contrarian indicators. When investors become overwhelmingly optimistic, there may be fewer incremental buyers left. When they become extremely pessimistic, much of the selling may already have occurred. There is logic to that argument. But sentiment is a notoriously imprecise timing tool.

The chart shows some earlier incidents when sentiment may have been misleading. For instance, bearish sentiment surged repeatedly during 2022 before the market finally established a durable low. The latest 53.3% reading is unusually pessimistic, but bearishness reached still higher levels during portions of 2022. So, we would be cautious about translating “investors are bearish” into “stocks must immediately rise.”

We can appreciate the sudden trepidation. The approaching midterm elections add uncertainty around the policy and regulatory landscape, while the Federal Reserve’s first rate hike in three years—and the prospect of additional tightening—has pushed short-term Treasury yields higher and increased the discount rates investors apply to equity valuations. The November 3 midterms are now less than two months away, and Fed policymakers have signaled the possibility of additional tightening. With markets still near historically elevated levels, that combination of political uncertainty and a less accommodative Fed could leave equities more vulnerable to volatility in the months ahead. A little skittishness is to be expected. But at the same time, the VIX index, which just closed at 14.81, doesn’t scream that investors are expecting near-term volatility.

Therefore, we remain circumspect. Rather than betting on timing or magnitude, we can simply say that psychology has changed—as the AI trade has changed. Investors are no longer approaching the market with the same certainty and blind faith that characterized much of the past few years.

The charts may be telling us two different things.

That distinction is probably the most interesting part of all this:

The AAII chart is about the broad market. It says investors have become unusually pessimistic.

Our AI basket comparison is about market leadership. It says the extraordinary relative advantage enjoyed by AI beneficiaries over companies perceived to be threatened by AI has reversed dramatically.

Those two signals aren’t necessarily loading on the same dimension. For instance, the broad equity market could stabilize as pessimism recedes while the former AI leaders continue to lag. The market could also broaden, with previously neglected companies performing better even as the major indices recover. Alternatively, the recent AI reversal could prove temporary, and the old leadership could reassert itself.

These charts, the substantial other data we analyze, and our intuition do not yet allow us to distinguish among those outcomes.

What they do suggest is that we should be careful about assuming that a contrarian signal from broad investor sentiment automatically tells us what to buy. Extreme bearishness may eventually become supportive of equities without necessarily restoring the exact leadership structure that existed before the decline. That would not be unusual because market cycles often turn before leadership cycles do.

The last shall be first and the first shall be last?

For several years, the AI investment debate has often been framed as a relatively simple division between winners and losers. Reality is likely to be messier.

Companies initially viewed as vulnerable to AI may adapt their businesses, use the technology to reduce costs, or discover that existing customer relationships, proprietary data, and distribution networks are more durable than expected. At the same time, companies clearly benefiting from AI can remain excellent businesses while producing less extraordinary investment returns if expectations and valuations have already moved far enough.

That is one of the enduring difficulties of thematic investing. Identifying an important technological change is not the same thing as knowing precisely which securities will capture the economic value from it—or at what price. The internet transformed the economy. So did mobile computing. The ultimate winners were not always the companies investors initially expected, and wonderful businesses were not necessarily wonderful investments at every valuation. Artificial intelligence is unlikely to be different in that respect.

The AI story is still being written.

There is a temptation when looking at sudden and acute reversals to declare the end of something. We think this story is more complex. Clearly, AI remains an extraordinary technological development. Trading volume, market capitalization, and the continued focus of the capital markets all point to AI remaining central to the investment landscape. But the market had also constructed a very specific expression around that story: own the beneficiaries and avoid—or short—the companies expected to be disrupted.

Episodes like the unraveling of Situational Awareness remind us that investing isn’t easy. The simple trade worked extraordinarily well. Now it has suffered a dramatic reversal. Perhaps that is the message in these charts?

At the same time, investor pessimism has risen sharply as one of the market’s most powerful leadership trades has unwound. That may eventually create opportunities. But the opportunity does not necessarily have to be simply buying yesterday’s winners at lower prices. It may instead be a broader market, a different set of winners, or simply a period in which the distinction between AI “beneficiaries” and AI “victims” becomes considerably less obvious.

We had recently noted that, from a technical perspective, the market may have become “oversold.” We see investor sentiment in the basement. Those are often contrarian indicators. The combination of an oversold tape plus very bearish sentiment can suggest an interesting entry point. But some of the deterioration in sentiment may be due more to disillusionment with one half of the AI trade than to feelings about the broader market. The reversal of the obvious trade should give investors a dose of humility—AI will likely remain a dominant economic and investment theme in the near term, but the winners and losers may not be as clear-cut.

Appendix

Basket constituents

Each basket holds 20 stocks, equally weighted. The lists are Belvest’s own and are intended as transparent proxies for the two themes, not replications of any third-party index.

Important disclosures