13F Season Landed (14 August 2026): Berkshire's Only New Position Was Worth $580,504 — and Its $10bn Alphabet Buy Was Announced 74 Days Earlier

The Q2 2026 Form 13F reports landed on Friday 14 August, and the most-read filing in the market is a near-perfect demonstration of what the form cannot do. Berkshire Hathaway reported a $299.25bn equity book across 29 issuers. Exactly one position in it was new: 3,564 shares of D R Horton, worth $580,504. Fifteen of the 29 showed an unchanged share count. And the headline the filing generated — a $37.76bn Alphabet stake, now the third-largest holding when both share classes are combined — rests substantially on a $10bn private placement that Alphabet itself disclosed to the SEC on 1 June, 74 days before the 13F was filed. The filings are stale by rule, incomplete by rule, and priced at a date that has already passed. Markets still move on them, and the reason is worth separating carefully from the reason people assume.

That gap between what a 13F reveals and what it merely repeats is the whole subject. A 13F does not put buying pressure on a stock. The buying it describes finished, at the latest, on 30 June. What arrives on disclosure day is a signal about a manager's stated reasoning, and any price move comes from participants who act on that signal now — not from the original trade, which the market absorbed six weeks ago. Netflix demonstrated the point a day before the deadline, and Berkshire demonstrated the limits of it a day after.

What actually happened: 29 positions, one of them new

Berkshire Hathaway's 13F-HR was filed on 14 August for the period ending 30 June 2026. It reports 89 line entries across 15 included managers, aggregating to 29 distinct issuers and $299,253,556,246. The March filing reported the same count of issuers and $263,095,703,570.

Set the dollar totals aside for a moment and count the decisions instead. One position entered the book: D R Horton, 3,564 shares, $580,504. One left it: Constellation Brands Class A, 632,890 shares, carried at $94.9m three months earlier. Fifteen of the 29 positions show a share count identical to the previous quarter. That leaves fourteen positions where anything at all was done, and the doing was concentrated in a handful of names.

The distinction between "new position" and "added to" is the one that gets lost fastest in a filing season, and it is not pedantic. Delta, Macy's, Lennar and both Alphabet classes were all in the 31 March filing. What the June filing records is a change in size, which is a different act with a different information content: adding to a position you already hold tells you about conviction, while opening one tells you about discovery. Only D R Horton is the second kind, and at $580,504 it is small enough that the rule would have permitted its omission entirely had the manager held fewer than 10,000 shares and under $200,000 in value — it clears one of those tests but not the other.

The $10bn that had already been announced

The Alphabet increase is the quarter's headline and the cleanest illustration in the whole filing of what a 13F adds and what it merely repeats.

Across both share classes Berkshire went from 57,835,013 shares worth $16.63bn to 105,979,600 worth $37.76bn — an 83% increase in shares, and 12.62% of the equity book. Combined, that is larger than the $32.51bn Coca-Cola stake and behind only Apple and American Express. The form does not present it that way, because Class A and Class C are separate securities with separate CUSIPs: they appear as the fourth and tenth largest lines, and the reader has to add them up.

But the larger point is the calendar. On 1 June 2026, Alphabet filed a free writing prospectus announcing an $80bn equity raise to fund AI compute infrastructure, and disclosed within it an agreement to sell $10bn of stock to Berkshire in a private placement — $5bn of Class A at $351.81 per share and $5bn of Class C at $348.20. At those stated prices the placement works out to roughly 14.2 million Class A shares and 14.4 million Class C. Berkshire's filing shows 24,541,369 Class A and 23,603,218 Class C added over the quarter. Subtract one from the other and roughly 19.6 million shares — about $7bn at 30 June prices — were acquired outside the pre-announced placement.

Note also where the most-quoted number of the week came from. The figure of roughly $19.8bn in net equity purchases during the quarter is not in the 13F and cannot be derived from it — a 13F reports holdings, not transactions, and contains no cash-flow statement. It comes from the 10-Q filed on 10 August, four days before the 13F: purchases of equity securities of $39,405m and sales of $27,780m for the six months to 30 June, against $15,938m and $24,087m for the three months to 31 March in the previous 10-Q. The difference gives second-quarter purchases of $23,467m against sales of $3,693m — net buying of $19,774m, following net sales of $8,149m in the first quarter. Two filings, two different questions, and only one of them answers "how much was actually spent".

What was actually due on 14 August

Section 13(f) of the Securities Exchange Act was adopted as part of the Securities Acts Amendments of 1975. The SEC's own description of the purpose is unusually direct: Congress wanted to increase the public availability of information on institutional holdings, believing the programme "would increase investor confidence in the integrity of the United States securities markets". Rule 13f-1, adopted in 1978, set the machinery — quarterly reports from any institutional investment manager exercising investment discretion over $100 million or more in Section 13(f) securities, each due within 45 days of the quarter end.

Three mechanical points follow, and each one is routinely misread.

First, 45 days after 30 June 2026 is 14 August 2026, a Friday. Under Rule 0-3(a), a deadline landing on a weekend or holiday moves to the next business day — the SEC's worked example is the fourth-quarter 2025 report, whose 14 February 2026 deadline fell on a Saturday before a Monday holiday and therefore shifted to Tuesday 17 February. No shift applied this quarter. The next one does apply: 45 days after 30 September 2026 is Saturday 14 November, so the third-quarter reports are due Monday 16 November.

Second, the deadline is a last permitted date, not a coordinated release. The filing window had been open since 1 July. Managers who filed early were visible early; those who filed at the close of business on the deadline day produced the concentrated burst of coverage that gives "13F season" its name — which is why a document describing 30 June was read as Friday-evening news.

Third, and most consequentially, the filing describes holdings "as of the end of the calendar quarter". Every share count and every dollar value in the reports published on 14 August is a 30 June figure. A position sold on 1 July still appears. A position built on 2 July does not.

Netflix, and what a six-week-old position is worth

The clearest live demonstration arrived a day before the deadline, and through a different document. On Thursday 13 August 2026, Netflix shares jumped nearly 4% after Pershing Square Capital Management disclosed a new position in its semiannual report — not a 13F, but subject to the same fundamental lag between the trade and the telling.

The stated reasoning was specific. Pershing said it had "acquired a position in Netflix, a business we briefly owned in 2022 and have followed closely ever since", and argued the company had "effectively won the streaming wars", citing more than 325 million subscribers against nearly half that for Disney+ and HBO Max combined, and noting that shares had fallen roughly 50% from their June 2025 high of $134, taking the valuation to about 21 times forward earnings from more than 40 times. Pershing's earlier involvement is part of the story: it built a large Netflix position in early 2022 and sold the entire stake about three months later, after the company reported its first subscriber decline in more than a decade. Neutral coverage: CNBC.

Hold the investment argument aside — pip theory takes no view on any security — and look only at the mechanism. Why does a price move on news of buying that already happened?

The four things Form 13F legally cannot show you

Each of these is a feature of the rule, not an evasion by any filer.

There is a fifth constraint that is less a blind spot than a units problem, and this quarter's filing measures it precisely. Because values are fixed at the quarter-end fair market value, the dollar figures have already been overtaken by whatever the market has done since 30 June — and, more immediately, the quarter-on-quarter change in a dollar figure is not evidence that anything was bought or sold. Ten of Berkshire's positions above $2bn had an identical share count in March and June. Their reported values did this:

Not one share changed hands in any of those lines. Occidental's stake appears to have shrunk by $4.35bn and Apple's to have grown by $8.11bn; both share counts are to the share identical. Berkshire's $36.16bn increase in reported value therefore has two sources mixed into a single number — actual purchases and market prices — and the form gives you no way to separate them from the dollar column alone. Share counts, not dollar values, are the stable quantity across quarters. Any read of a 13F that begins by comparing dollar totals is measuring the market's opinion of the holdings, not the manager's.

The currency question the data cannot answer

There is a recurring temptation to read 13F aggregates as evidence of foreign appetite for US assets, and therefore for the dollar. The filings cannot support that reading, for a structural reason: Section 13(f) reporting is confined to securities on the Official List, which is a list of US exchange-traded instruments. A foreign manager crossing the $100 million threshold files its US equity book and nothing else — its domestic holdings, its bonds and its currency hedges are all outside the perimeter. A 13F therefore shows the size of a manager's US equity exposure, but never the exposure it chose instead, which is precisely the comparison a cross-border capital-flow story requires. That comparison lives in balance-of-payments data, not in EDGAR. For the factors that actually drive the dollar in the meter's framework — rates, growth, positioning, risk sentiment and commodities — the USD currency page carries the live read.

The $100 million line, and why it has not moved since 1978

The threshold is the most quietly consequential number in the whole regime, and it is nearly five decades old.

In July 2020 the SEC proposed raising it to $3.5 billion. The reasoning in the proposing release was arithmetic: the $100 million figure came from the 1975 statute and represented a certain proportion of the US equity market at that time, and since the form's 1978 adoption, US public corporate equities had grown from $1.1 trillion to $35.6 trillion. A $3.5 billion threshold would restore roughly the original proportion. The Commission estimated it would retain over 90% of the dollar value of reported holdings while relieving nearly 90% of filers, and noted that the aggregate value reported under the existing threshold represented approximately 75% of the US equity market, against 40% in 1981.

The proposal was never adopted. The SEC's current staff guidance still states the $100 million threshold, which is the operative fact for anyone filing or reading a filing this week. The episode is instructive for a different reason: in setting out the costs of the regime, the SEC explicitly identified front-running and copycatting of smaller managers' portfolios as consequences of publication. A disclosure rule that is officially recognised to change how others trade is not a neutral window onto the market — it is part of the market.

What this does, and does not do, to the instruments you trade

The realistic scope of 13F season is single-name volatility on disclosure days, concentrated in liquid large caps that attract attention because of who owns them. The index-level effect is close to nil, and the reason is the same lag that makes the data stale: whatever the aggregate institutional book looked like on 30 June, it was already expressed in prices during Q2. Nothing is added to US500 or NAS100 demand by describing it in August.

That distinguishes 13F season sharply from the scheduled events that genuinely reprice an index — a mega-cap earnings report, where the information is about the future and arrives simultaneously for everyone. The Home Depot preview for 18 August is the contrast case: guidance changes estimates, and estimates change index-level earnings. A 13F changes neither. Where holdings data does interact with a live storyline is around lock-ups and newly listed names, where institutional ownership is genuinely still forming — the SpaceX lock-up expiry is the clearer example of position data mattering because the positions are new, not because they are disclosed.

The takeaway

Form 13F is a 1975 disclosure statute doing 2026 work. It reports one side of one asset class, at prices that are six weeks old, from managers above a threshold that has not been adjusted since 1978, with legal permission to omit small positions and to defer sensitive ones. Every one of those limits is knowable in advance, which is the useful part: a reader who knows what the form excludes cannot be surprised by what the form excludes.

This quarter supplied the demonstration in a single filing. The most closely read 13F in the market recorded one new position worth $580,504, fifteen untouched share counts, a $36.16bn swing in reported value driven substantially by prices rather than trades, and a headline stake whose largest component had been announced by the issuer 74 days earlier. None of that means the filing was uninformative — the residual Alphabet accumulation, the 58% cut to Capital One, the 142% add to Macy's are all real, and all genuinely new on 14 August. It means the informative part was a subset of the document, and identifying which subset requires reading the previous quarter's filing beside it and knowing which other filings got there first.

The Netflix session was the model for the week, and the mechanism held. Those shares moved not because a large buyer was buying — that finished in the second quarter — but because a stated argument became public and other participants acted on it. Whether any such argument turns out to be right is a separate question from whether the disclosure mechanism works, and the disclosure mechanism is the only part of this that is fully documented. Read the filings for what they are: a legally defined, deliberately partial record of the past, published late, and priced by people reacting in the present.

More on how pip theory frames drivers rather than forecasts on the about page.

Educational macro context only — not investment advice.