What the best 13F funds are losing on, and what they are up eightfold on

For every position in a fund's top 50 the database carries an estimated average purchase price. Divide the June 30 price by that estimate and each position has a gain or loss versus cost. Across the 24,746 positions at top-rated funds, 16.1% are below cost, the median position is up 44%, and 31% have more than doubled. The middle half runs from +9% to +128%. 5.6% are down 20% or more.

The bottom-rated funds' 82,836 positions look almost the same: 17.1% under water, median +43%, 29% doubled, middle half +8% to +118%. Two things separate the groups, and neither is the loss rate. The top-rated funds' upper tail is fatter (a 75th percentile of +128% against +118%), and, as the scorecard showed, the top-rated funds sized those winners larger.

In dollars, $173.7bn of the top-rated funds' $2.70tn in top-50 positions sits below cost, 6.4%. The dollar share is lower than the position share because the losing positions are, on average, the smaller ones.

Sort the top-rated funds' positions by their weight in the book. Positions under 1% of the book: 20.5% under water, median gain +31%. One to three percent: 13.6% and +48%. Three to five: 11.1% and +70%. Five to ten: 9.9% and +85%. Over ten percent of the book: 9.8% under water and a median gain of +110%, with 52% of those positions more than doubled.

Part of this is arithmetic: a position that has tripled is, by that fact, a larger share of the book than it was when bought, so large weights and large gains go together. Part of it is not. The under-1% band holds the positions funds have been trimming, the new positions bought recently at recent prices, and the ones that fell. Either way, for a reader looking at a 13F, weight is the most useful single number on the page: a top-rated fund's 10% position was, at June 30, almost never a loss.

Restrict the view to stocks held by at least ten top-rated funds and take the median holder's gain versus cost. 84 of the 635 widely held stocks had a median holder below cost at June 30 (at least eight holders). The deepest, among names with ten or more holders: Intuit (−34%, 20 holders), Boston Scientific (−33%, 20 holders, $2.8bn), Salesforce (−28%, 45 holders, $2.8bn), then Pfizer, Adobe, ServiceNow (46 holders), Coupang, Alibaba (39 holders), Disney (39), T-Mobile, Medtronic, PepsiCo and CME.

The list has a shape. Application software (Intuit, Salesforce, Adobe, ServiceNow) is the largest cluster; medical devices (Boston Scientific, Medtronic) and payments (PayPal, Fiserv, Global Payments) follow; the rest are a mix of defensive consumer and Chinese e-commerce. These are the stocks the top-rated funds bought and were, at June 30, wrong about so far.

The second dot on the chart moves the comparison to the September 24, 2026 close. Salesforce (+11% now), ServiceNow (+11%), CME, AT&T, Danaher and Abbott have since crossed back above the median holder's cost. Intuit (−29%), Boston Scientific (−29%), Coupang (—) and PepsiCo have moved further below it. The count of widely held stocks under water rose from 84 at June 30 to 118 at the latest close, because the summer took more from the winners than it gave to the losers.

The other end of the table is the AI hardware trade held from early. Micron: 243 top-rated holders, $96.0bn of positions, median holder up 695% on cost at June 30 and 633% at the latest close. Vicor (+472%), KLA (+461%, 114 holders, $40.3bn), Sandisk, Lam Research, Sterling Infrastructure, Erasca, Corning and Western Digital follow, all with the median holder up three- to fivefold.

Between June 30 and September 24, 2026, most of those gains narrowed. KLA's median holder went from +461% to +249%, Lam Research from +415% to +260%, Applied Materials from +307% to +163%, Corning from +360% to +189%. AMD is the exception, up further. The flow study found the same funds selling these names into the June quarter's rally; the cost data says why they could afford to.

By industry, medical devices lead with 56% of top-rated positions below cost and a median gain of −3%; packaged foods 52%; application software 49% across 920 positions and $66.7bn, the largest pool of dollars on the list; exchanges and data providers 49%; telecom services and medical instruments 47%. By sector, telecommunications, consumer staples, health care and consumer discretionary all sit above the 16% average; technology (12% under water, median +92%) and industrials (10%, median +55%) sit well below it.

The positions that matter most are the ones a fund has sized to matter. 88 positions at investable top-rated funds (at least $250m disclosed, eight or more names) were 5% or more of the book and 10% or more below cost at June 30. The largest by weight: Deer Management with 35% of its book in StubHub at −24% versus cost, now −69%; Himalaya Capital with 22% in PDD at −22%; Greenoaks with 19% in Coupang at −45%, now −55%; Voyager Global with 17% in Microsoft at −14%, since recovered to +15%; Paulson with 14% in Bausch Health at −60%; Apollo with 12% in Rackspace at −66%, now −80%.

Some of these are not stock picks in the usual sense: Apollo's Rackspace and Bain's Surgery Partners are private-equity sponsor stakes held through a listing, and StubHub is a recent IPO whose cost is its offer price. The rest are the contrarian book of the top-rated funds as of June 30: Chinese e-commerce (PDD, Alibaba, Coupang's Korean equivalent), a handful of software names bought on the way down, and Microsoft, which had one of the largest funds on the list under water at mid-year and above cost by September.

For stocks with at least eight holders in both groups, compare the median purchase price of top-rated holders with that of bottom-rated holders. Across 563 such stocks the crowd paid more in 46% of cases and the median gap is −0.4%: on the typical stock, the two groups paid about the same. The tail is one-sided. On Cogent Biosciences the crowd's median cost is 163% above the elite's ($25 against $10); on Sandisk 106% ($900 against $436); on Ciena, Vicor, Lumentum, Sterling Infrastructure and nLight between 80% and 100%. Every name on the list is a stock that ran, and in every case the bottom-rated funds' cost sits closer to the top.

This is the flow study's finding seen from the other side. In Q2 the bottom-rated funds were net buyers of the memory and optical names the top-rated funds were selling; here, their average entry into those same names sits well above the elite's. The top-rated funds' edge in these stocks was not owning them, it was owning them first.

Search the 635 stocks held by at least eight top-rated funds. Gains are the median holder's gain versus estimated cost at June 30 and at the September 24, 2026 close; the split-adjusted latest price is used where a split occurred after June 30.