The Stock Market’s AI Winners Are Not Who You Expected

A June 2026 note from JPMorgan’s U.S. equity quant team found that retail stock pickers beat Wall Street benchmarks in single stocks this year by concentrating on the same names powering the AI trade. The research, from strategist Arun Jain, is the latest data point in the AI stock market story of 2026, and it cuts against decades of evidence showing stock picking is a loser’s game. Everyday investors tilted heavily into Micron, AMD, and Nvidia, and their picks trounced strategies that dollar-cost average into the Nasdaq 100.

The same week, Nasdaq announced that CoreWeave, Nebius, Astera Labs, Teradyne, and Rocket Lab will join the Nasdaq 100 on June 22, the latest quarterly rebalance that affects about $1.4 trillion in tracked assets. The new entrants are almost entirely AI infrastructure plays, the same names retail has been loading up on, and the same companies whose AI capex is now reshaping the broader economy. Citadel Securities puts that AI capex at 2% of U.S. GDP, or roughly $650 billion, with about 2,800 data centers planned for construction.

The Surprise Winners Sitting Across From Wall Street

The retail beating Wall Street story sounds too clean to be true, but the math is on Jain’s side. ‘In single stocks, retail has unsurprisingly outperformed benchmarks over the past month or so, consistent with a concentrated tilt toward MU, AMD, and NVDA,’ Jain wrote, per Sherwood News. The same note found retail beating the S&P 500 year to date, and only lagging the Nasdaq 100 because that index itself has been lifted by the same AI names retail chose directly.

• 89% of global trading volume is now AI-driven (LiquidityFinder)
• 60% to 70% of trades are now executed algorithmically (LSE Research)
• 79% of U.S. large-cap equity fund managers trailed the S&P 500 in 2025 (S&P Dow Jones Indices)
• $650 billion in AI capex in 2026, or 2% of U.S. GDP (Citadel Securities)
• ~2,800 data centers planned in the U.S. (Citadel Securities)

Two of those numbers carry the story. The first, 89%, is the share of global trading volume that AI now drives. The second, 79%, is the share of large-cap equity fund managers who trailed the S&P 500 in 2025. Together they explain why a research note from one quant strategist has reset the retail-versus-Wall-Street conversation for the year.

JPMorgan’s data lines up with S&P Dow Jones Indices’ finding that 79% of large-cap equity fund managers underperformed the S&P 500 in 2025, the kind of stat that powers the long-running joke about active management. Yet in 2026, everyday investors with the smallest research budgets have built portfolios that beat the pros on the most-watched stocks in the market. The names that did the work, Micron, AMD, and Nvidia, are the same ones on every bank’s restricted list, the same ones in every quant fund’s crowded trade.

When 15,000 AI Agents Read the Same Script

The opposite edge of the same trade showed up three months earlier, in what the DeFi community now calls the ‘February Wick.’ In a single three-second window in February 2026, $400 million in liquidity vanished from a popular Solana liquidity pool. The cause was not a hack or a whale but an estimated 15,000 autonomous AI trading agents running variations of the same open-source strategy, all triggering the same sell signal at the same block.

1. Feb. 3: Anthropic ships new AI productivity tools. Nasdaq falls 1.4%, S&P 500 0.8%, Dow 0.3%. Data services, software, and software-focused private-equity firms lead the slide (the Feb. 3, 2026 selloff after Anthropic’s AI tools).
2. February: The ‘February Wick.’ 15,000 AI trading agents crash a DeFi market in three seconds, draining a popular Solana liquidity pool. Aave processes $180 million in liquidations with zero bad debt (the breakdown of the 3-second AI-agent crash).
3. April: Anthropic releases Claude Design. Figma and Adobe slide. Adobe is down nearly half over 12 months by early June.
4. June 9: JPMorgan publishes a note showing retail traders beating Wall Street benchmarks in single stocks (the JPMorgan note on retail traders beating benchmarks).
5. June 10: Super Micro prices a $7 billion equity offering to fund $39 billion in AI server orders, sending AI stocks lower (Super Micro’s $7B raise triggering an AI stock sell-off).
6. June 22: CoreWeave, Nebius, Astera Labs, Teradyne, and Rocket Lab join the Nasdaq 100.

The pattern, in the analysis by BlockEden.xyz, is a problem the 2010 Flash Crash first flagged. When trading systems share a strategy, they also share a blind spot. ‘When machines trade in perfect synchronization, they don’t distribute risk, they concentrate it into a single point of catastrophic failure,’ the post-mortem read. Aave’s automated liquidation system processed $180 million in collateral in under ten seconds, with zero bad debt, a quiet vindication of DeFi’s no-circuit-breaker design, and a louder warning about what synchronized agents can do in less time than a human can read a chart.

Speed matters, and the bigger factor is sameness, the convergence of strategies across the agent fleet. The February Wick’s block at 1,234,570 closed in three seconds, with the order book reloading before a human trader could read the candle, and the speed is the new floor, not the ceiling.

Scale, plus synchronization, plus speed, is what made the February Wick uniquely destructive. The same dynamic has a counterpart in the labor market, where Citadel Securities noted software engineer job postings are up 11% year over year and the unemployment rate is 4.28%. AI is replacing workers on the S-curve the technology industry has always followed, not the steep curve the headlines suggest.

The market risk runs in a different direction from the labor story. It is the same condition the SEC’s Feb. 3 speech tried to name without using the word ‘synchronized.’ Both phrases describe a market that has stopped behaving like a population of independent actors and is starting to behave like a single trader with a single view of the world.

What the SEC Sees as AI Leaves the Loop

U.S. regulators have noticed. On February 3, the same day the Anthropic tools hit the tape, the director of the SEC’s Division of Investment Management gave a speech in Florida (the SEC’s February 2026 speech on AI in investment management) arguing that AI represents a generational shift for the industry. The agency’s response so far has been to ask questions rather than draft rules.

AI is here. And intelligent use of artificial intelligence can, should, and will catalyze a transformation of the technology of investment management.

The same speech, delivered in Florida by the SEC’s director of the Division of Investment Management, floated a specific pilot: replacing the traditional fund prospectus with a fund-trained AI agent that an investor could query in plain English. The director’s broader point was that the gap between what AI can do and how Wall Street is deploying it is now a human problem inside large organizations, not a technical one.

The research world’s view is harsher. Maximilian Goehmann, a PhD candidate at the London School of Economics, told a UK Treasury inquiry that algorithmic systems already dominate order flow, and that small data errors can cascade into market-wide events. ‘There were a lot of algorithms with similar settings that were each triggering each other,’ he said of the 2010 Flash Crash. ‘One feedback loop was triggering the next, which triggered the next, and that created a cascading failure.’ His proposed fix is not more rules but voluntary data certification that lets market participants signal quality to win trust, rather than imposing it.

The SEC’s speech and the LSE research line up on one point. The next market shock is unlikely to come from a single rogue algorithm. It is more likely to come from a thousand well-behaved algorithms behaving the same way at the same time, the same shape of risk that took $400 million out of a Solana pool in three seconds in February.

The Cost on Wall Street Is Showing Up in Jobs and Stocks

The losers from the same shift are easier to see in stocks than in headcount. Adobe, the company at the center of the creative-software stack, posted a fiscal Q2 revenue beat of $6.62 billion on June 11, raised its full-year guidance, and still saw its shares fall 5.5% in after-hours trading. The stock is down nearly half over the past 12 months, pressured by the April release of Anthropic’s Claude Design, a generative design product aimed at the same customers. The CEO who ran Adobe for 18 years, Shantanu Narayen, is on his way out, and the CFO, Dan Durn, is set to step down next week.

February 3 gave the selloff its first big test. The Nasdaq fell 1.4%, the S&P 500 0.8%, and the Dow 0.3% in a single session as data and software stocks sold off on fears of AI displacement. Thomson Reuters, FactSet, S&P Global, and Intuit led the slide, while private-fund managers with software exposure, Ares Management and Blue Owl Capital, both dropped about 10%.

The session showed that the AI trade has a sharp downside, and that the consensus on whose business is at risk changes month by month. The New York Times reported in April that AI is now eliminating jobs across Wall Street, with the headcount cuts running through attrition rather than the layoffs that defined the post-2008 industry. Adobe’s quiet transition, with its CEO and CFO both leaving in the same quarter, is the visible version of the same dynamic. The same companies that powered the indexes in 2024 are the ones being de-rated in 2026, and the de-rating is happening faster than the index providers can rebalance around it.

The Quiet Winners Behind the AI Trade

The clearest beneficiaries of the AI trade sit underneath the trading desks, in the chip and cloud companies that supply the compute the agents run on. The Nasdaq’s June 22 rebalance is the latest data point on the shift. Five of the five new entrants to the Nasdaq 100 are AI infrastructure plays, including two neocloud GPU providers in CoreWeave and Nebius, and three hardware specialists in Astera Labs, Teradyne, and Rocket Lab.

The chip side of the table is the most aggressive bet. Citi double-upgraded AMD to Buy in June with a $575 price target, up from $460, on the view that AMD’s custom GPU business will scale to $33 billion in 2027 sales, up 137% year over year.

Memory is following the same curve. The price of DRAM that cost a Baltimore IT shop $100 six months ago now costs $300, a shift big enough that the Federal Reserve is tracking it as an inflation input (AI memory chip prices as a US inflation problem). The application layer tells the same story. Stripe’s ‘Payments Foundation Model’ boosted fraud detection on card-testing attacks to 97% from 59%, per AI Street (transformer models in finance and the 27.8% accuracy breakthrough), with the same publication reporting that a transformer-based model called LOBERT hit 27.8% accuracy in next-message prediction, against 6.1% for the previous leading model.

The infrastructure winners are a different set of names than the consensus thought. They are the neoclouds renting GPUs to the hedge funds, the foundries building the GPUs for the neoclouds, and the memory vendors supplying the foundries. India’s $47 billion AI trade is following the same pattern across a different set of industrial names (India’s $47 billion AI trade hidden in industrial stocks), and the AI trade is paying them a rent tied to the same $650 billion capex plan, with another 2,800 data centers on the way.

Where the Synchronized Risk Lives

The same infrastructure that lifted retail and the chip vendors is what makes the market more fragile. Goehmann’s research suggests 60% to 70% of all trades are now executed algorithmically, a number that has climbed as AI deployment has spread from quant funds to retail platforms. When the strategies converge, the market behaves like a single trader with a single view of the world. Speed matters, and the bigger factor is sameness, the convergence of strategies across the agent fleet.

Scaling model size and expanding data coverage offer a promising path toward improving predictive performance in asset-return forecasting. However, progress remains constrained by computational limitations and the scarcity of large-scale, high-quality financial data.

The Citadel Securities note made a quieter point about labor markets that complicates the doomer case. AI diffusion in the labor market is, on the data so far, not the steep curve the headlines predicted. Software engineer job postings are up 11% year over year, and the unemployment rate is 4.28%.

The remaining risk is the synchronized kind, the one regulators say is hardest to measure. When the same model is in production across the agent fleet, the diversity that used to come from human disagreement is gone, and the fix, per the SEC’s speech, the LSE research, and the academic quoted in AI Street, is more data, more compute, and more variation in the strategies that get deployed, not more rules.

Frequently Asked Questions

What did JPMorgan find about retail traders in 2026?

JPMorgan strategist Arun Jain, the firm’s head of U.S. equity quant strategy, found that retail stock pickers beat dollar-cost-averaging into the Nasdaq 100 in single stocks. The outperformance was driven by a concentrated tilt toward AI names including Micron, AMD, and Nvidia. The same note found retail beating the S&P 500 year to date.

What was the February Wick?

In February 2026, an estimated 15,000 AI trading agents running variations of the same open-source model sold the same assets at the same block on a Solana liquidity pool. The event is now known as the February Wick, a single three-second candlestick that wiped about $400 million in liquidity. Aave processed $180 million in liquidations with zero bad debt in the same window.

Is the SEC regulating AI trading?

Not with new rules yet. The SEC’s Feb. 3, 2026 speech by the director of the Division of Investment Management asked the industry for pilot proposals and floated the idea of replacing the traditional fund prospectus with a fund-trained AI agent that can answer investor questions in plain English. The agency is calling for engagement, not enforcement.

Are Wall Street jobs being lost to AI?

Yes. The New York Times reported in April that AI is now eliminating jobs across Wall Street, with the headcount cuts running through attrition rather than the layoffs that defined the post-2008 industry. Adobe is the latest visible example, with both its long-serving CEO, Shantanu Narayen, and its CFO, Dan Durn, set to leave in the same quarter.

Which stocks have been the biggest AI winners in 2026?

On the retail side, JPMorgan points to Micron, AMD, and Nvidia. On the index side, Nasdaq added CoreWeave, Nebius, Astera Labs, Teradyne, and Rocket Lab to the Nasdaq 100 effective June 22, the latest quarterly rebalance that affects about $1.4 trillion in tracked assets.

Disclaimer: This article is for informational purposes only and is not financial advice. Stock prices, price targets, and the cited research figures are accurate as of publication. AI trading carries synchronization and concentration risks described in the body. Consult a qualified financial professional before making investment decisions.