Q2 13Fs: Hedge Funds Rotated Within AI Instead of Leaving It

When the Q2 2026 13F filings landed, the easy headline was "hedge funds sell Nvidia." The more useful reading is that managers kept their AI exposure but moved it around. Capital flowed out of some prior winners and into new issuers, hyperscalers and infrastructure names. For anyone using 13F data, that distinction matters, because a rotation and a retreat call for very different conclusions.

What the filings suggest

Roundups of the second-quarter filings describe a pattern of broadening rather than wholesale de-risking. One manager-by-manager wrap-up of the quarter characterised it as managers rotating the trade instead of simply adding more of the same winners. Other summaries of the filings note hedge funds increasing positions in SpaceX and Alphabet while reducing Nvidia and Broadcom, with several named funds exiting Nvidia outright, including D1 Capital, Discovery Capital and Third Point.

At the same time, large-cap AI names showed up on both sides of the ledger. Nvidia, Taiwan Semiconductor and Alphabet were added by some managers and reduced by others. Amazon was added by funds such as Appaloosa, and trimmed by others such as Bridgewater. That is what a rotation looks like in aggregate: no single direction, but a lot of repositioning underneath.

New names are absorbing the flow

The clearest sign of broadening is how many funds opened positions in newly public companies. SpaceX appeared as a new position for a large number of filers, and Cerebras showed up as a new holding for a meaningful group as well (a rough read of the filings, not a rigorous tally). We looked at the former in our piece on SpaceX as the most-opened position of Q2 and at ownership of the latter in who owns Cerebras.

New listings pull money from somewhere. When a fund opens a large position in a freshly listed company, it often funds that purchase by trimming a position that has already worked, which can explain part of the selling in older semiconductor leaders without implying a bearish view of AI demand.

Why selling a winner is not a bearish call

Three things can sit behind a reduced stake in a name like Nvidia:

1. Position-size discipline. After a large run, a holding can outgrow a fund's risk limits, and trimming restores the original weight. See our note on portfolio weight versus share count.

2. Funding a new idea. Cash raised from one AI name may go straight into another.

3. Hedging changes the picture. A 13F shows long equity positions only, so some trims are offset by options or short exposure we cannot see. We covered this in what 13Fs cannot show.

Our earlier look at Nvidia in the Q2 filings found that passive and market-making filers can dominate the headline share changes, which is one more reason to look at who is doing the buying and selling.

How to test the rotation thesis yourself

If you want to check whether a rotation is happening, rather than assume it, a few steps on InvestorLens help:

1. Open the flow page and compare net buyers and net sellers across several AI-linked names at once. A true retreat shows sellers everywhere, while a rotation shows buyers and sellers in the same theme.

2. Use the overlap tool to see which new AI holdings are shared by many top investors, and whether they also hold the older leaders.

3. Browse investor profiles to separate long-horizon holders trimming for discipline from funds that exited entirely.

4. Check the macro consensus page to see whether tech's overall share of institutional capital actually fell.

Caveats that matter

The Q2 snapshot is as of June 30, so everything above is already more than four months old. Markets have moved since, and the 45-day filing delay means fund managers may have changed course again; our guide to using the 13F lag explains how to handle that. Crowding is also a real risk when many funds chase the same new issues, as we discussed in crowded trades and overlap warning signs.

The takeaway

Q2 13F data does not show smart money walking away from AI. It shows funds rebalancing within the theme: taking gains in older leaders, adding hyperscalers, and opening positions in new listings. The real test comes with the Q3 filings due by mid-November. If the same new names appear again as buys and the trimmed leaders keep losing holders, the rotation is a trend. If not, Q2 may simply have been a one-quarter reshuffle.

---

Data sourced from public SEC 13F filings. Educational research only — not investment advice.