Reviewed November 6, 2026. This study compares public Form 13F positions for the quarters ended March 31 and June 30, 2026.
Which widely held stocks divided the same 17 institutional reporting groups most sharply in Q2 2026? Amazon (AMZN) produced the broadest directional split: among 11 managers that reported Amazon common stock in both quarters, five disclosed more shares and six disclosed fewer—the largest opposing camps in the screen. Alphabet (GOOGL/GOOG) followed with five increasers, four decreasers and two mixed class-level cases; Taiwan Semiconductor Manufacturing (TSM) had four increasers and seven decreasers. The result is not a vote, a consensus forecast or a reconstruction of trades. It is a direction count from public quarter-end share quantities.
The disagreement was broad, not an isolated Amazon result. Of the 20 issuers held continuously by at least four groups in our fixed panel, 19 had at least one manager reporting more shares and at least one reporting fewer. Lam Research was the exception: four of five continuous holders reported fewer shares and one was unchanged. A pooled total can therefore hide a materially split manager distribution.
The answer in one table
| Issuer | Continuous holders | More shares | Fewer shares | Unchanged | Mixed classes |
|---|---|---|---|---|---|
| Amazon (AMZN) | 11 | 5 | 6 | 0 | 0 |
| Alphabet (GOOGL/GOOG) | 11 | 5 | 4 | 0 | 2 |
| TSMC (TSM) | 11 | 4 | 7 | 0 | 0 |
| Meta Platforms (META) | 10 | 3 | 7 | 0 | 0 |
| NVIDIA (NVDA) | 9 | 3 | 6 | 0 | 0 |
| Microsoft (MSFT) | 8 | 3 | 5 | 0 | 0 |
| Broadcom (AVGO) | 7 | 3 | 4 | 0 | 0 |
| AMD (AMD) | 6 | 2 | 4 | 0 | 0 |
| DoorDash (DASH) | 5 | 2 | 2 | 1 | 0 |
| Uber (UBER) | 5 | 2 | 2 | 1 | 0 |
| ASML (ASML) | 5 | 3 | 2 | 0 | 0 |
| Micron (MU) | 5 | 3 | 2 | 0 | 0 |
Amazon shows why the pooled total is incomplete
The five Amazon increasers—Altimeter, Appaloosa, ARK, Coatue and Whale Rock—reported a combined 17,515,046 shares in Q1 and 23,299,059 in Q2, an increase of 33.0%. The six decreasers—Bridgewater, Light Street, Pershing Square, Soros Fund Management, Third Point and Tiger Global—went from 29,851,281 to 23,256,425 shares, a decline of 22.1%.
Across those 11 continuous holders, the pooled count slipped only 1.7%, from 47,366,327 to 46,555,484 shares. That small net move does not mean the panel was stable. It is the arithmetic result of two large, opposing groups. Nor is the combined count unique ownership: the same issuer is reported separately by multiple managers, and the sum is used only to describe this fixed comparison panel.
Alphabet’s two classes require a mixed label
Alphabet demonstrates why issuer-level totals cannot always be forced into a single direction. Berkshire Hathaway increased both disclosed classes. Bridgewater decreased both. Coatue increased one class and decreased the other, so we classify it as mixed. ARK increased its continuing class and also reported a second class in Q2, which we also keep as mixed rather than treating the issuer-level sum as an unqualified increase.
This treatment is deliberately conservative. It prevents a class addition, removal or internal class shift from being misread as a clean directional call on the issuer. The same principle applies to amendments, linked reporting entities and corporate-action exceptions: unresolved structure is isolated instead of being filled with an invented adjustment.
Technology exposure was not one coordinated trade
The divergence extends across large technology and semiconductor names. TSMC split 4–7, NVIDIA 3–6, Microsoft 3–5 and Broadcom 3–4. Our prior AMD-versus-NVIDIA continuous-holder study showed that rising quarter-end reported values could coexist with lower aggregate share counts. The present study answers a different question: how many managers moved each way.
That distinction matters because breadth, weight and direction are not interchangeable. The holder-breadth versus portfolio-weight analysis found that more reporting holders did not always mean more pooled weight. The pairwise overlap study likewise showed that shared names and shared capital weights can produce different leaders. Here, even the same continuous-holder subset can split in opposite share-count directions.
Method: 35 source tables, 17 reporting groups, class-aware directions
- Panel: Berkshire Hathaway; Pershing Square; Dalal Street; Appaloosa; Greenlight/DME; Fundsmith; Third Point; the Icahn reporting group; Soros Fund Management; Bridgewater; ARK; Tiger Global; Coatue; Altimeter; Whale Rock; Situational Awareness; and Light Street.
- Periods: quarter ends March 31 and June 30, 2026. We used 35 official information tables because Pershing Square’s Q1 comparison requires two complementary reporting entities.
- Eligible scope: operating-company common equity, ordinary shares, ADRs, REITs and operating-partnership equity. Funds, options, debt or principal positions, warrants, preferred/depositary equity, commodity trusts and dated equity units are excluded.
- Continuous holder: the reporting group disclosed the issuer in both quarters. New Q2 names and Q1-only names are not converted into “increase” or “decrease.”
- Direction: as-filed SH share quantities are compared by common-stock CUSIP. Multiple rows for the same manager, security and quarter are combined. If classes move in opposite directions, or one class appears while another continues, the manager-issuer observation is mixed.
- Corporate actions: no separate split factor was applied to the headline single-class issuers; their CUSIP and SH share-unit definitions were comparable across the two filings. This is an as-filed quantity comparison, not a claim that identifier continuity alone proves the absence of every possible corporate action.
- Screen: the main ranking requires at least four continuous holders. Twenty issuers qualified; 19 had both increasers and decreasers.
The SEC states that Form 13F information tables report issuer, security class, shares or principal amount and quarter-end fair market value. Multiple lines may exist for one issuer because classes or investment-discretion categories differ. That is why we normalize by reporting group and security before classifying direction.
What this analysis cannot show
Form 13F does not disclose actual trade dates, purchase prices, manager intent, short positions, non-13F assets or complete firm-wide exposure. A higher quarter-end share count is not proof that every share was bought during the quarter; transfers, reporting-perimeter changes and corporate actions can affect comparisons. A lower count is not a performance result. This study therefore describes disclosed direction within a defined public universe and should not be treated as investment advice or a ranking of managers.
For the panel definition, individual manager histories and other original statistics, see the AHNIM Studio Guru hub.
Three findings to retain
- Amazon had the broadest directional split: five continuous holders increased shares and six decreased, the largest opposing camps in the screen.
- Nineteen of 20 issuers with at least four continuous holders had managers moving in both directions; pooled totals conceal that dispersion.
- Alphabet required two mixed labels because class-level directions differed, showing why issuer totals need security-level reconciliation.
Sources and original documents
Accessed November 5–6, 2026. Primary source links are direct SEC EDGAR documents. The SEC guidance pages explain the required fields and reporting scope; AHNIM Studio owns the eligibility rules, issuer reconciliation and calculations.
- SEC, Frequently Asked Questions About Form 13F
- SEC, Create an XML Information Table for Form 13F
- Altimeter: Q1; Q2
- Appaloosa: Q1; Q2
- ARK: Q1; Q2
- Berkshire Hathaway: Q1; Q2
- Bridgewater: Q1; Q2
- Coatue: Q1; Q2
- Dalal Street: Q1; Q2
- Greenlight / DME: Q1; Q2
- Fundsmith: Q1; Q2
- Icahn reporting group: Q1; Q2
- Light Street: Q1; Q2
- Pershing Square: Q1 entity 1; Q1 entity 2; Q2
- Situational Awareness: Q1; Q2
- Soros Fund Management: Q1; Q2
- Third Point: Q1; Q2
- Tiger Global: Q1; Q2
- Whale Rock: Q1; Q2