Fund Managers Are Underweight AI by a Record Margin. That Could Be the Market's Next Tailwind

Key Takeaways

• Record AI underweight. Large-cap mutual funds are about 1.75 percentage points underweight AI-exposed stocks versus their benchmarks, the widest gap on record. This figure excludes the mega-caps Amazon, Broadcom, Alphabet, Meta, Microsoft and Nvidia.
• A full reversal in two years. In mid-2024, funds were overweight AI names by about 0.40 percentage points. Today AI stocks make up 13.0% of the average large-cap fund versus 14.8% of benchmarks, up from roughly 6.5% for both in Q3 2024.
• The benchmark outran the portfolios. AI stocks' share of the benchmark more than doubled. Funds added exposure but could not keep pace .
• The underweight extends to the biggest names. Funds are about 100 bps underweight Nvidia, 70 bps underweight Alphabet, 60 bps underweight AMD and 50 bps underweight Microsoft 
• It is costing performance. Fidelity funds have named underweights in AMD, Lam Research, Applied Materials and KLA among their largest detractors (Fidelity).
• Institutions are raising cash. Global fund manager cash rose 0.4 points to 3.9%, the biggest monthly jump since March. Equity overweights fell from 56% to 49% of managers .
• Bonds remain deeply unloved. A net 48% of managers are underweight bonds, the most since May 2022 and the 17th straight month of underweight positioning (Halifax / BofA FMS).
• The top fear has shifted from AI to yields. A disorderly rise in bond yields (33%) has replaced an AI bubble (28%) as the biggest tail risk (Halifax / BofA FMS).
• The underweight could become buying pressure. If AI earnings keep delivering, underweight managers face pressure to chase benchmark weights, and higher cash levels give them room to buy.

 

Overview

AI has been the defining market story of the past two years, yet America's large-cap mutual fund managers own less of it, relative to their benchmarks, than at any point on record. That gap could turn into buying pressure if AI keeps delivering.

Large-cap mutual funds are now about 1.75 percentage points underweight AI-exposed stocks versus their benchmarks, the largest underweight on record. That figure excludes the mega-caps: Amazon, Broadcom, Alphabet, Meta, Microsoft and Nvidia. AI-exposed equities make up 13.0% of the average large-cap fund portfolio, compared with 14.8% of the benchmarks those funds are measured against. In mid-2024 the picture was reversed: funds were overweight AI names by about 0.40 percentage points. In Q3 2024, both the fund weight and the benchmark weight stood near 6.5%.

In roughly two years, AI stocks' share of the benchmark more than doubled. Fund managers added exposure too, but not fast enough to keep up.

How the Gap Opened

Goldman Sachs' positioning research describes the problem directly. The weight of AI infrastructure stocks in mutual fund portfolios "has risen sharply this year," but it "has failed to keep pace with benchmark weights, leaving mutual funds significantly underweight the complex" (24/7 Wall St., citing Goldman Sachs). Hedge funds are the opposite case: their returns have tracked swings in the AI trade closely in recent months.
 

The underweight also runs through the biggest names:

The gap isn't new either. A year ago, Goldman found funds were 107 bps underweight a basket of AI names excluding the Magnificent 7. Their underweight to the Magnificent 7 itself had reached 819 bps, and Information Technology sat at a record 536 bps underweight (Investing.com).

Why Managers Are Underweight

Several forces are working together:

1. Benchmarks moved faster than portfolios. When a group of stocks doubles its index weight in two years, an active manager has to buy aggressively just to stay neutral. Doing that means selling other holdings the manager still believes in, often at a time when AI valuations look stretched. Many managers chose to rebalance slowly, and the gap grew.

2. Concentration limits. Diversified mutual funds operate under diversification rules and internal position caps. As the largest AI names grow to take up a big share of an index, it becomes structurally difficult to match their weights, and managers often end up underweight by default.

3. Valuation discipline and bubble worries. Professional investors are openly uneasy about how much is being spent on AI. In BofA's September Global Fund Manager Survey, a net 33% of respondents said companies are overinvesting, matching the record high set in February. The share naming AI hyperscaler capex as the most likely source of a systemic credit event rose to 42% from 38% (Halifax / BofA FMS).

4. The trade already feels crowded. "Long global semiconductors" is the most crowded trade in the BofA survey, cited by 53% of respondents (Investing.com). When a trade feels crowded, managers hesitate to add, even when their benchmark keeps pushing that exposure higher.

5. Scars from the July shakeout. The Philadelphia Semiconductor Index fell nearly 29% from its June 22 high to its July 29 low, then rebounded about 20% in roughly three weeks (Moomoo / Zhitong Finance). De-risking across tech and AI accelerated through June and July, and many funds had not rebuilt their positions ahead of Nvidia's earnings (Crypto Briefing).

The Cost Shows Up in Performance

These underweights have hurt returns, and fund managers' own commentaries say so:

Meanwhile, Institutions Are Raising Cash

The AI underweight comes as professional investors pull back on risk more broadly. In BofA's September survey, global managers' cash allocation rose 0.4 percentage points to 3.9%, the biggest monthly increase since March. In early August it was 3.5%, the sixth-lowest reading since the survey began in 1998 (The Kobeissi Letter). The global portion of the survey covered 170 participants overseeing $470 billion. The full panel was 190 managers with $512 billion (CANSLIM Blog).

On Treasury policy, nearly 50% of managers expect the Treasury's buyback program to have no effect on yields, while only 16% expect it to lower them (The Kobeissi Letter, Investing.com).

In other words, investors are locking in gains after the global equity rally. Bank of America's Michael Hartnett noted it is only "safe to increase risk exposure when cash back in 4-5% neutral zone," so the survey's cash rule is still flashing caution (Investing.com).

Why the Underweight Could Become a Tailwind

A record underweight does not mean managers are bearish or short AI. It means they hold less than their benchmark. That distinction matters because active managers are judged on relative performance. If AI stocks keep outperforming, underweight managers face tracking error, pressure on relative returns and career risk. Each of those pushes them toward "benchmark-chasing" purchases to close the gap (Moomoo / Zhitong Finance).

Early signs of that are already showing. Goldman prime-brokerage data show institutions raised their net allocation to the Magnificent 7 from about 14% at the end of June to about 18% in August. TMT stocks drew the largest net buying in the U.S. market, driven mainly by new long positions rather than short-covering (Moomoo / Zhitong Finance).

There are two ways this can play out:

 

The Bottom Line

The professionals who manage America's large-cap mutual funds are underweight AI by the widest margin on record. That's not because they reject the theme. The benchmark doubled its AI exposure faster than portfolios could follow, and concentration limits, valuation concerns and a crowded-trade backdrop all made catching up harder. At the same time, global managers are raising cash and cutting equity overweights. Investors aren't abandoning the AI trade; they're banking gains and waiting for a clearer entry point.

That setup is what could make the underweight a tailwind. When professional money is this far below benchmark on a theme that keeps delivering earnings, even a modest push back toward neutral weight could become a steady source of buying pressure for AI stocks.

Tickeron AI Perspective

 Disclaimers and Limitations