The latest 13F filings reveal a notable, though limited, shift in institutional investors’ behavior toward mega-cap technology stocks and AI-related sectors. While institutions have not abandoned their technology bets, second-quarter 2026 data indicate that they have become more selective in allocating capital, particularly toward the “Magnificent Seven,” which have led the market’s rise in recent years.

A Reuters analysis of 13F filings submitted to the U.S. Securities and Exchange Commission (SEC) found that approximately 6,371 investment firms disclosed their positions through the end of June. The analysis showed that about 44% of these firms reduced their holdings in the “Magnificent Seven,” while 42% increased their holdings or opened new positions. The firms included pension funds, hedge funds, wealth managers, and other institutional investors. The figures point to greater disagreement among portfolio managers, but they do not by themselves reveal the direction of net cash flows, since they compare the number of firms that increased or decreased their holdings rather than the value of the money moving between them.

These figures do not necessarily mean that institutions expect a collapse in mega-cap technology stocks. Rather, they reflect a shift from broad enthusiasm toward more precise management of risk, valuations, and exposure to crowded trades—that is, bets in which a large number of investors are positioned in the same direction.

Greater Caution Toward the “Magnificent Seven”

The “Magnificent Seven” includes some of the largest U.S. technology companies, including Microsoft, Meta, Alphabet, Amazon, Nvidia, and other giants that have come to represent a significant weight in institutional portfolios.

The importance of the 13F figures lies in showing that the market is no longer moving on the basis of a unified institutional bet on this group. The gap between firms that reduced their positions and those that increased them is relatively narrow, suggesting that investors differ significantly over whether current prices truly reflect these companies’ expected future growth.

Some analysts believe the explanation may be simpler than a change in their view of the companies themselves: some institutions may have already reached the limits permitted by their portfolios’ risk and concentration policies. When a stock represents a large share of a portfolio, an investment manager may have little room to increase the allocation, even if they remain convinced of the company’s long-term fundamentals.

Therefore, not buying more of a stock does not necessarily mean a desire to sell it; it may simply reflect risk-management constraints or portfolio concentration limits.

Lack of Consensus Does Not Mean Abandoning AI

This distinction appears even more important when looking at AI stocks and related infrastructure.

While institutions have become more cautious toward some mega-cap technology stocks, they have continued to show buying appetite for a group of AI-related stocks. According to the Reuters analysis, approximately 36% of the investment firms were net buyers of the AI-related stocks covered in the analysis, including companies such as Arista Networks and Broadcom.

However, the nature of this demand appears different from the early stages of the AI wave. Some investors have begun viewing these stocks as momentum trades highly sensitive to valuations, liquidity, and leverage, rather than solely as long-term investments supported by earnings growth.

This highlights an important point: institutional caution does not equal rejection of AI. It may instead mean that investors believe growth in the sector will continue but have become more selective about which companies they want to own and how much they are willing to pay for that growth.

Semiconductors Remain Attractive

The picture is more positive in the semiconductor sector. An analysis of second-quarter filings showed that approximately 48% of investment firms were net buyers of sector stocks, compared with 34.5% that were net sellers. This means the institutional bias remained positive, even as risk management became a more prominent factor in investors’ decisions.

This finding is important because it prevents an exaggerated conclusion from being drawn from the 13F data—that institutions are “fleeing technology.” The reality is more complex: investors reduced some positions in mega-cap technology stocks, but they did not abandon the infrastructure underpinning the AI boom.

The data also showed divergence among sectors tied to this wave. Interest in data-center stocks and some areas adjacent to AI was more balanced, while institutions were more inclined to sell certain software and energy stocks.

Tiger Global Offers a Clear Example

There may be no clearer example of position resetting than the moves made by Tiger Global Management during the second quarter.

According to the fund’s filings, Tiger Global reduced its Alphabet holding by 45.4% to 5.81 million shares through the end of June. It also cut its Nvidia position by 6.8% to 11.20 million shares. It reduced its Microsoft holding by 9.3% to 2.27 million shares, its Amazon holding by 3.2% to 9.68 million shares, and its Meta holding by 8.5% to 2.82 million shares.

The cuts were not limited to the “Magnificent Seven.” Tiger Global reduced its Broadcom position by about 51% to 1.75 million shares and cut its holding in Taiwan Semiconductor Manufacturing Company by 12.3% to 4.88 million shares.

But the other side of the story is equally important. The fund increased its Intel holding from 1.64 million to 4.25 million shares, more than doubling its position from the previous quarter. It also established a new position in AMD totaling 674,727 shares, along with a 375,000-share position in SpaceX.

Tiger Global’s moves therefore do not look like a wholesale exit from technology, but rather a reallocation within the sector itself: reducing exposure to some names whose valuations had risen sharply while increasing exposure to others, in a move that can be interpreted as a search for opportunities with different valuation profiles or growth potential.

When Crowded Trades Become a Problem

After the quarter ended, July’s turbulence exposed the risks of crowded technology trades.

According to a JPMorgan report cited by Reuters, global hedge fund managers gave up about 3% of their cumulative gains since the beginning of the year in July as technology-related trades were unwound. Despite July’s losses, global hedge funds across strategies remained up about 8% year to date, according to JPMorgan. The report noted that investors’ crowding into specific technology bets made it more difficult to exit positions when the market trend reversed.

These developments highlight a familiar problem in markets: when a large number of investors bet on the same direction, entering the trade may be easy during an upswing, but exiting becomes more difficult when sentiment changes simultaneously.

Additional signs of this shift emerged in July, as hedge funds increased their short bets on certain AI stocks, according to Hazeltree data reported by Reuters.

13F Filings Are Not a Complete Picture of the Market

Even so, these data should be treated with caution. A 13F filing is not a real-time disclosure of an investor’s portfolio.

The Securities and Exchange Commission states that institutional investment managers overseeing at least $100 million in securities covered under Section 13(f) must file quarterly disclosures within 45 days of the end of the quarter. For the second quarter of 2026, the deadline was August 14, 2026.

More importantly, the 13F form shows eligible holdings as they stood at the end of the quarter and does not necessarily reflect what an investor owns when the filing is published. Short positions also do not appear on the 13F and are not deducted from long positions.

Therefore, an institution reducing its position in a stock during the second quarter does not necessarily mean it had turned bearish on the stock in August. Likewise, increasing a position does not mean the institution still held the same position after the end of June.

Repositioning Rather Than Retreat

Overall, second-quarter filings do not provide sufficient evidence that the “Magnificent Seven” have peaked or that institutions expect the AI wave to end.

They do, however, offer a more nuanced signal: indiscriminate buying of technology appears to be more difficult than it was before.

Institutions are still buying semiconductor stocks, and appetite remains for AI-related companies, but at the same time they are reducing some positions in the largest technology companies and reallocating capital among different names and sectors.

From a risk-management perspective, this shift may be closer to portfolio rebalancing than to a signal that expectations for technology are collapsing. After years of money flowing toward a limited number of companies, investors appear to be asking a more difficult question: not whether AI will transform the economy, but which companies will actually capture the largest share of the profits generated by that transformation, and at what price should their shares be bought?

If this shift continues, the next chapter of technology’s rise may depend less on the strength of the “Magnificent Seven” as a single group and more on investors’ ability to distinguish between companies with genuine earnings growth and those benefiting primarily from market momentum.