Back to Writing

DYOR

Institutional and Corporate Position Changes: 13F Plus Schedules 13D/13G

A combined SEC review of 13F and Schedule 13D/13G filings covering major banks, famous managers and strategic corporate holders. The report corrects the timing of NVIDIA's 9.3% Nebius beneficial ownership disclosure and compares Q2 early-filer consensus.

Bottom line: the highest-information corporate ownership disclosure in the latest 30 days is NVIDIA's 9.3% beneficial ownership of Nebius. However, the SEC filing says most of it comes from a pre-funded warrant acquired on March 11—not a purchase on July 13. After stock-split and corporate-action adjustments, NVDA, JNJ and SNOW were net adds for five of the six Q2 early filers.

The timing distinction is essential. Schedules 13D/13G disclose beneficial ownership events such as crossing a 5% threshold and are generally more timely; 13F compares quarter-end portfolios. This report uses both but keeps them in separate tables and never treats event date, filing date and acquisition date as interchangeable.

Treat this as an idea-generation map, not an automatic buy list. Current valuation, earnings revisions, technical entry points and short exposure are outside the filing data. The sensible next step is to research high-consensus names now and rerun the screen after the full August filing deadline.

SEC EDGAR · 2026-07-21 · FACT / INFERENCE SEPARATED

Filed in the last month does not mean bought in the last month

After stock-split and corporate-action adjustments, NVDA, JNJ and SNOW were net adds for five of six comparable Q2 early filers. AI infrastructure and defensive assets appeared together rather than as a one-factor risk-on trade.

1,74513F-HR and amendments filed in 30 days
1,588unique filer CIKs
6comparable managers in the core sample
$221.7bncombined disclosed Q2 value
Critical limitation: the Q2 2026 13F deadline is August 14. The June 22–July 21 pool contained 1,701 original 13F-HR filings and 44 amendments; 1,563 had a June 30 report date, but this remains a heavily biased early-filer sample.
Correction: this chart update adds corporate-action adjustments. After KLA's 10-for-1 split, KLAC falls from an unadjusted 6/6 to 2/6 and about $55.35m. After Booking's 25-for-1 split, BKNG falls from 4/6 to 3/6 and about $122.58m. Both leave the top-12 screen; the prior conclusions are withdrawn.

Focus: strategic and large beneficial ownership from Schedules 13D/13G

The 30-day SEC index contained 1,312 unique 13D/13G accessions, including 456 initial, non-amended filings. The table below curates high-information cases involving well-known corporations or a major bank. Schedule 13G generally denotes passive ownership, while Schedule 13D may involve control or governance; neither automatically proves an open-market purchase.

Filer → issuerFiled / event dateBeneficial ownershipWhat it actually meansSignal
NVIDIA → Nebius (NBIS)
Schedule 13G
Jul 20 / Jul 1322,256,412 shares · 9.3%1,190,476 shares were already in NVIDIA's Q1 13F; the other 21,065,936 are underlying a pre-funded warrant acquired March 11. The warrant entered the 60-day beneficial-ownership window on July 13. This was not a 22.3m-share purchase on July 13.High strategic value
timing needs correction
Palantir → Surf Air Mobility (SRFM)
Schedule 13G
Jul 1 / Jun 248,248,989 shares · 7.4%An initial 13G confirms passive beneficial ownership above 5%. The filing does not provide cost basis, and the event date alone does not prove a same-day purchase.Strategic watch
Renesas → SiTime (SITM)
Schedule 13D
Jul 9 / Jul 13,558,691 shares · 11.9%Renesas sold its timing business to SiTime for about $1.5bn cash plus these shares and gained a board nomination right. The stake was transaction consideration, not open-market buying.High industrial logic
JPMorgan → ALGM / MKSI / KLIC / VICR
four Schedules 13G
Jul 16 / Jun 30ALGM 6.8%; MKSI 5.0%; KLIC 5.0%; VICR 6.9%The bank and its accounts crossed passive 5% ownership thresholds. This is useful position-size evidence, but it is not a JPMorgan corporate strategic investment or proof of concentrated July buying.Institutional ownership
Correction to the screenshot: labeling July 13 as “Purchase / New Position” is too simplistic. NVIDIA's original SEC 13G says the warrant was acquired on March 11, 2026; July 13 was when the September 11 exercise date came within the 60-day beneficial-ownership window. This is a high-value signal that NVIDIA's relationship with Nebius is meaningful, but not evidence of a new 22.3m-share July trade.

Q2 early-filer coverage: largest disclosed 13F portfolios

We read tableValueTotal from the SEC cover page for all 1,563 Q2 filings and deduplicated amendments to 1,552 managers. These figures are disclosed 13F security values, not firm AUM. Security-level comparisons then use representative managers with usable adjacent-quarter records.

RankManagerFiled13F valueRows
1SwedbankJul 15$115.99bn1,123
2AssenagonJul 8$80.18bn1,158
3Wealth Enhancement AdvisoryJul 15$59.79bn4,384
4Ninety One UKJul 16$44.73bn271
5Envestnet Portfolio SolutionsJul 20$20.87bn1,369
6Pennsylvania Public School ERSJul 17$20.59bn1,759
7Davenport & CoJul 13$20.35bn1,356
8Sequoia Financial AdvisorsJul 17$20.07bn2,163
9Fort WashingtonJul 17$19.89bn452
10Czech National BankJul 8$18.35bn503

Visual analysis: inspect the sample before the consensus

The figures use a research-style sample flow, lollipop plot and categorical heatmap. They are descriptive statistics. Because the core sample is not random, we do not report population confidence intervals or significance tests.

Figure 1 · Sample-selection flow

Core sample ≈ 0.39% of Q2 managers
Figure 1 note: this is a data-quality filter, not random sampling. Figures 2–3 should not be extrapolated into a population-wide probability of buying.

Figure 2 · Top 12: consensus breadth versus estimated added notional

line length = estimated added notionalright column = net-add managers
Figure 2 note: the top 12 are selected by breadth and then estimated notional, and plotted in descending notional order. Twenty-nine securities reached at least 4/6; only the top 12 are shown. Notional uses quarter-end implied prices and is not execution value.

Figure 3 · Manager-by-security action heatmap

Add ANew NReduce RNot held ·
SecuritySwedbankAssenagonCongressATPAvanzaMUFG-MSTotal
NVDAAAARAA5
JNJAAAAAR5
SNOWAAN·AN5
GOOGLAARRAA4
AVGOAAR·AA4
PEPAA·AA·4
WMTAAR·AA4
GEVAAR·AA4
MUAANRAR4
PLTRAAA·AR4
PGAARAAR4
NEERAAAAR4
Figure 3 note: A is add, N is new, R is reduce, and a dot is not held. Rows are ordered by breadth and then estimated notional; a reduction does not imply a short position.
High breadth and notional, but concentrated: NVDA reached 5/6 and about $1.24bn, yet roughly 94.5% of estimated added notional came from Assenagon.
A more balanced 5/6: JNJ totaled about $413m, with Assenagon and Swedbank contributing roughly 55% and 43%, respectively.
High breadth, low notional: SNOW appears at 5/6 managers, including two initiations, but only about $125m—idea diffusion rather than heavy conviction.
Corporate actions are first-line QC: split adjustment moves KLAC from 6/6 to 2/6 and BKNG from 4/6 to 3/6; skipping this step reverses the ranking.

1. Strongest cross-manager signals

Direction is based only on reported share-count changes in ordinary long equity positions. Estimated added notional equals added shares multiplied by quarter-end reported value per share. It is a proxy, not execution value or cost basis.

RankSecurityCUSIPNet-add managersEstimated added notionalInterpretation
1NVDA · NVIDIA67066G1045 / 6~$1.24bnStrong breadth and notional; highly concentrated in Assenagon
2JNJ · Johnson & Johnson4781601045 / 6~$413mDefensive more balanced five-manager consensus
3SNOW · Snowflake8334451095 / 6~$125mTwo initiations; high breadth but lower notional
4GOOGL · Alphabet A02079K3054 / 6~$789mPlatform AI exposure
5AVGO · Broadcom11135F1014 / 6~$566mCustom silicon and networking
6PEP · PepsiCo7134481084 / 6~$462mDefensive staples exposure
7WMT · Walmart9311421034 / 6~$457mDefensive quality retail
8GEV · GE Vernova36828A1014 / 6~$365mPower and data-center infrastructure
9MU · Micron5951121034 / 6~$287mMemory; new at Congress
10PLTR · Palantir69608A1084 / 6~$206mAI software and government data
11PG · Procter & Gamble7427181094 / 6~$153mDefensive staples exposure
12NEE · NextEra Energy65339F1014 / 6~$89mDefensive utility and power demand

2. What each core manager changed

Manager / filedQ2 disclosed valueRaw New / AddRepresentative adjusted initiations or addsQuality note
Swedbank
Jul 15
$115.99bn44 / 326SpaceX new $609m; AMD +$521mHoldings report; 2 other managers; KLAC is a reduction after split adjustment
Assenagon
Jul 8
$80.18bn176 / 461SNDK new $1.41bn; NVDA +$1.17bn; UBS +$808mLarge multi-strategy portfolio
Congress Asset Management
Jul 7
$16.25bn32 / 150ARM new $145m; KLIC new $117m; MIR +$100mNormal early Q2 holdings report
ATP
Jul 17
$4.60bn18 / 107DAL new $32.2m; MRK new $30.6m; MTB new $30.2mDanish supplementary pension fund; KLAC is a reduction after split adjustment
Avanza Fonder
Jul 17
$4.59bn13 / 246MSFT +$33.3m; AMZN +$31.1m; AAPL +$28.3mSubstantial indexed/fund exposure; KLAC is a reduction after split adjustment
Mitsubishi UFJ Morgan Stanley Securities
Jul 15
$32.7m10 / 12AMAT +$3.2m; AVGO +$2.2m; MRVL new $0.82mSmall portfolio; used as direction confirmation

Raw New / Add counts are not fully corporate-action-adjusted and only describe turnover; they are not a buy count. Leading securities in the main screen and figures were checked for splits and identifiable CUSIP migrations. ETFs, convertibles and options still require security-level review.

3. Major banks, brokers and famous managers: Q2 still pending

Fact: as of July 21, the major institutions below had not filed their main Q2 2026 13F. Any “what they bought” data in this section compares their latest available Q1 2026 holdings with Q4 2025 and is not mixed into the Q2 early-filer screen. Bank portfolios include client, market-making, custody and multi-strategy accounts, so single-security changes carry less manager-signal content than concentrated funds.
Major bank / brokerLatest main 13FLargest Q1 net adds vs restated Q4Q2 statusInterpretation note
JPMorgan Chase2026Q1 · May 13 restatementGOOG +$5.25bn; META +$3.39bn; MRK +$2.81bnPendingRestatement replaces original filing
Goldman Sachs2026Q1 · May 15MSFT +$3.63bn; NVDA +$2.46bn; GOOG +$2.36bnPendingHigh turnover and client exposure
Morgan Stanley2026Q1 · May 15AAPL +$3.55bn; NVDA +$3.35bn; AMZN +$2.57bnPendingCombination; Q4 uses later restatement
Bank of America2026Q1 · May 18JPM +$2.63bn; META +$1.53bn; BMY +$1.06bnPendingConsolidated firm and client accounts
Citigroup2026Q1 · May 11NVDA +$2.26bn; MSFT +$1.79bn; AAPL +$1.59bnPendingNot suitable for copy-trading
UBS Group2026Q1 · May 5MSFT +$1.85bn; AAPL +$1.39bn; TSM +$1.26bnPendingCombination; large client component
Nomura2026Q1 · May 15TECK new $377m; CTSH +$275m; WBD +$132mPendingCombination report

Bank comparability QC: JPMorgan and Morgan Stanley had relatively stable reporting boundaries. Included-manager counts changed at Goldman, Citi, UBS and Nomura, while Bank of America's table-row count fell sharply. False AstraZeneca initiations caused by a CUSIP migration were removed. All estimates describe consolidated firm/client quarter-end positions, not proprietary trading.

Famous investors and managers: representative latest-Q1 moves

ManagerLatest filingLargest Q1 New / AddQ2 status
Berkshire HathawayMay 15 · $263.1bnGOOGL +$10.47bn; DAL new $2.65bn; GOOG new $1.03bnPending
BridgewaterMay 15 · $22.4bnAMZN +$508m; TSM new $364m; AVGO +$207mPending
Pershing SquareMay 15 · $13.7bnMSFT new $2.09bn; AMZN +$384mPending
AppaloosaMay 15 · $5.9bnAMZN +$446m; UBER +$322m; SNDK new $179mPending
Soros Fund ManagementMay 15 · $9.1bnLIN new $129m; EA +$110m; SEE +$94mPending
Tiger GlobalMay 15 · $22.8bnTSM +$622m; AMAT +$260m; MELI new $233mPending
CoatueMay 15 · $29.1bnEQIX new $1.07bn; ASML new $655m; TSM +$221mPending
Citadel / Millennium / D. E. Shaw / RenaissanceMay 14–15 · 2026Q1Included in the watch universe, but excluded from a copy-trade list because of high turnover and multi-strategy/quant structurePending
Scion Asset ManagementLatest: 2025Q3 · Nov 3, 2025No 2026Q1 main 13F; old positions cannot be assumed to remain currentNo recent filing

Well-known corporate investment filers

Corporate filerLatest disclosed valueRepresentative changeQuarter status
Japan Post HoldingsQ2 · $5.97bnOnly AFL; reduced by 1,076,010 shares, no New/AddQ2 filed
GlencoreQ2 · $4.87bnBunge and Century Aluminum share counts unchanged, no New/AddQ2 filed
SoftBank GroupQ1 · $11.4bnEthos Technologies new ~$34.9mQ2 pending
Markel GroupQ1 · $11.9bnSunbelt Rentals new $62.7m; MELI new $9.4mQ2 pending
LoewsQ1 · $11.9bnTEAM +$43.7m; BLDR +$37.9m; AMZN new $29.2mQ2 pending

The “well-known company” category only includes corporate entities that actually have a Form 13F filing obligation and a comparable securities portfolio. Famous public companies such as Apple or Tesla do not automatically file 13F simply because they are well known.

4. Portfolio message: offense and defense coexist

The broadest consensus spans offense and defense

NVDA, JNJ and SNOW reached 5/6. GOOGL, AVGO, MU and COHR reached 4/6 alongside defensive PEP, WMT and PG. The early sample is not a one-factor AI or risk-on basket.

Equal votes do not mean equal capital

MUMSS reports only about $32.7m versus Swedbank's roughly $116bn portfolio, yet each manager receives one breadth vote. Breadth must be paired with estimated notional and manager concentration.

Initiations deserve extra diligence

Congress initiated ARM, KLIC and MU; Assenagon and Avanza initiated SNDK; multiple managers initiated SNOW. Corporate actions, CUSIP changes and reporting-scope changes still need to be ruled out.

The complete answer arrives after mid-August

Many large managers file close to the deadline. This is a first-pass radar, and the consensus screen should be rerun after August 14.

5. How to use the screen

Research priority: start with NVDA, while noting that about 94.5% of its estimated added notional came from Assenagon; then compare valuation and earnings momentum across GOOGL / AVGO / MU / COHR; use JNJ / WMT / PEP / PG as defensive controls. SNOW belongs on an idea-diffusion watchlist, but its 5/6 breadth corresponds to only about $125m. Do the fundamental and valuation work first, then wait for a price trigger.
  • Good for: discovering institutional consensus, finding initiations, building sector watchlists and validating durable themes.
  • Not good for: timing entries, inferring actual cost basis, observing equity shorts or estimating net exposure.
  • Next update: rerun after the August 14 deadline and add valuation, estimate revisions and price performance.

6. Method and safeguards

StepMethodReason
Filing universeSEC master index; Jun 22–Jul 21; 13F-HR and 13F-HR/ASeparates filing date from report date
Core sampleOriginal Jun 30 13F-HR filings with an adjacent Mar 31 filingExcludes old-period amendments and delayed disclosures
Aggregation keyCUSIP + put/call + SH/PRNAvoids issuer-name and class-text drift
DirectionNew/Add/Reduce/Exit based on ordinary-long share countsMarket-value changes include price movement
Corporate actionsAdjust stock splits and map verifiable CUSIP migrations before comparing quartersPrevents mechanical share-count changes from becoming false trades
OptionsExcluded from the main ranking13F lacks strike, expiry, cost and delta
Quality controlFlags amendments, confidentiality, combination reports and other managersScope changes can create false buys
Open risks: stock splits, mergers, exchanges and CUSIP changes can create false initiations or exits. 13F omits equity shorts, cash and most bonds, and it does not necessarily represent a manager's proprietary view. Every “buy” here is strictly a net increase between two quarter-end snapshots.

Data cutoff: July 21, 2026. Facts come from original SEC filings; theme synthesis, added-notional estimates and research priorities are YoloLab inferences. Not investment advice.