Shortly after four o’clock on Friday afternoon, a document that an entire industry had been waiting on for weeks appeared on the SEC’s servers. Berkshire Hathaway disclosed its equity holdings as of June 30, 2026. It was the second full quarterly filing under Greg Abel, and it followed a quarter in which the company did something it had not done in three and a half years: buy more stock than it sold. Purchases came to $23.5 billion, sales to $3.7 billion. For the first time in fifteen quarters, Berkshire was a net buyer.
Open the filing and you will find exactly one new position in it. It is D.R. Horton. It consists of 3,564 shares and is worth $580,504. The total reported portfolio stands at $299.25 billion. For every dollar Berkshire holds in U.S. equities, roughly two millionths of a dollar sits in D.R. Horton.
The gap between $23.5 billion of buying and half a million dollars of new names is neither a contradiction nor a mystery. It is the most precise available description of what a Form 13F is and what it is not. And because thousands of investors spend this week reading the filings of famous money managers in order to derive trade ideas from them, it is worth working this one case through to the end.
What the document actually says
Berkshire reports 29 positions spread across 89 table rows and fourteen additional included managers, from National Indemnity to GEICO to Nebraska Furniture Mart. Total value rose from $263.10 billion in the first quarter to $299.25 billion. The position count stayed at 29. The filing is signed by Marc D. Hamburg, Senior Vice President, dated August 14.
The most revealing figure is not printed in the table; it has to be computed from it. Fifteen of the 29 positions are unchanged to the individual share versus the prior quarter: Apple at 227,917,808 shares, American Express at 151,610,700, Coca-Cola at exactly 400,000,000, plus Chevron, Occidental, Chubb, Moody’s, Kraft Heinz, SiriusXM, VeriSign, both Liberty Live classes, Louisiana-Pacific, NVR and Jefferies. Together those fifteen are worth $215.2 billion. Roughly 72 percent of the reported portfolio did not move by a single share all quarter.
On the sell side there is exactly one complete exit: Constellation Brands, 632,890 shares, gone. Beyond that, trims, some of them large. Capital One down 58 percent to three million shares, Nucor down 52.5 percent, Kroger down 22 percent to 39 million shares, Ally Financial down almost seven percent, DaVita down just over four. Apart from Alphabet, the additions were Delta Air Lines, up 44 percent to 57.32 million shares, Macy’s up 142 percent off a very small base, Lennar up nearly 30 percent, and The New York Times up 3.7 percent.
The only decision of the quarter is called Alphabet
Alphabet shows up twice in the filing because Berkshire holds both classes. The voting Class A stake went from 54,249,798 to 78,791,167 shares, an increase of 24,541,369 shares or 45.2 percent. The non-voting Class C stake jumped from 3,585,215 to 27,188,433 shares, up 23,603,218 shares or 658 percent. Combined: from 57.84 million to 105.98 million shares, an increase of 48.14 million shares or 83 percent. Value rose from $16.63 billion to $37.76 billion.
Part of that was already public. On June 1, Alphabet announced equity offerings totaling $80 billion to fund its buildout of AI compute infrastructure. Berkshire subscribed to a $10 billion private placement within it: $5 billion of Class A at $351.81 per share and $5 billion of Class C at $348.20. That works out to 14,212,216 Class A shares and 14,359,563 Class C shares, about 28.57 million shares in total.
Now the subtraction that did not appear in any press release. The stake grew by 48.14 million shares; the placement supplied 28.57 million. That leaves 19.57 million Alphabet shares Berkshire simply bought in the open market during the second quarter — 10.33 million of the A class and 9.24 million of the C class. At prices in the neighborhood of the placement price, that is roughly $6.5 to $7 billion. The exact figure cannot be derived from the filing, because 13Fs carry no cost basis. The order of magnitude can.
Which produces the real shape of the quarter. Of $23.5 billion in equity purchases, about $16.8 billion went into a single company. Roughly 72 percent of all buying went into one name — and almost exactly 72 percent of the portfolio never moved. Berkshire’s first expansive quarter in three and a half years consists, at its core, of one decision. Everything else is trimming.
Incidentally, that makes Alphabet the company’s third-largest U.S. equity position, behind Apple at $65.95 billion and American Express at $51.28 billion but ahead of Coca-Cola at $32.51 billion. Plenty of automated trackers list it fourth, because they count the two share classes separately. That is the most harmless of the errors this document invites.
Two positions that look new and are not
The less harmless error lives in the issuer name column. In the first quarter Berkshire reported a position under the label Bank America Corp. In the second quarter the same line reads Bank of Amer Corp. Anyone reconciling the two filings by name sees a complete exit from a $27 billion position and a brand-new $27 billion position. The CUSIP in both cases is 060505104. It is the same stock. What actually happened is the opposite of an initiation: the stake fell from 513,624,165 to 483,394,015 shares, a decline of 30.23 million shares or 5.9 percent.
The same mechanism hits Chubb. Last quarter Chubb Ltd Switz, this quarter Chubb Limited, CUSIP H1467J104 both times, share count exactly 34,249,183 both times. A name-based comparison reports an $11.67 billion sale and an $11.67 billion purchase. Nothing happened.
This is not pedantry. It is why headlines about institutional filings so often disagree with each other. Reconcile 13Fs by CUSIP, never by issuer name. The name column is free text typed by the filing manager, and it changes whenever somebody in Omaha updates a template.
What a 13F structurally cannot show
The obligation applies to institutional managers with at least $100 million in certain U.S. securities. What gets reported is long positions in exchange-listed U.S. equities, depositary receipts and some options and convertibles. What does not get reported: bonds, cash, stakes in private companies, short positions, hedges, and every share held on a foreign exchange.
That last exclusion produces the biggest omission in this particular case. Berkshire owns roughly ten percent of each of the five Japanese trading houses — Mitsubishi, Mitsui, Marubeni, Itochu and Sumitomo. Those stakes appear in no 13F and never will, because they trade in Tokyo. Greg Abel traveled to Japan that same week to meet the leadership of those exact five companies. Anyone tracking Berkshire’s portfolio purely through the American form does not see that part of the company at all.
Then there is time. The filing describes positions as of June 30 and was permitted to be submitted through August 14. Anything bought on May 15 and sold on June 20 never existed as far as the document is concerned. Berkshire filed on August 14 — the last permissible day. That is not unusual: research on filing behavior shows hedge funds push the deadline considerably harder than other managers, with only about a quarter filing by day 40 while roughly thirty percent file on day 45 alone. The motive most consistent with the evidence is protection of the filer’s own still-running trades.
Finally, a manager may omit individual positions entirely with the regulator’s permission. Berkshire has used that tool before, disclosing its Chubb accumulation only months after the fact — the same Chubb position now sitting untouched in the portfolio. In this filing, however, the field for confidentially omitted holdings is explicitly set to no. Nothing is missing this time.
Same afternoon, same stock, opposite direction
That same Friday, Bill Ackman’s Pershing Square disclosed that it had sold its remaining Alphabet holdings in both classes during the second quarter, after cutting 95 percent of the stake in the prior quarter. It added to Microsoft and Meta instead and trimmed Amazon.
Two of the most closely watched addresses in American finance, the same stock, the same quarter, opposite directions, published within the same hour. Read 13Fs as signals and you must conclude that a fundamental disagreement about Alphabet is being fought out here.
It probably is not. Ackman has said publicly that he still sees upside in Alphabet, but that his capital is limited and Microsoft currently looks like the better use of it. Berkshire, meanwhile, sits on $344.1 billion in cash and short-term Treasury bills. The two firms are not making the same decision under different assumptions; they are making different decisions under completely different constraints. A concentrated fund with a dozen positions has to sell in order to buy. Berkshire does not. The form shows the same kind of line in both cases — and in both cases omits the one circumstance that explains the line.
The same exercise works on Delta Air Lines, which Berkshire increased by 44 percent while other prominent managers cleared the position out over the course of the year. Here too: the document contains the action, never the reason.
$344 billion and a new man at the top
The context for this filing is the leadership transition. Under Greg Abel the cash mountain has shrunk for the first time in years, by roughly nine billion to $344.1 billion in cash and Treasury bills; a broader definition puts it near $365.5 billion. Operating earnings for the quarter came in at $13.0 billion, investment gains at $12.7 billion.
In parallel, Berkshire is buying its own stock again. On March 4 the company resumed repurchases after a nearly two-year pause with a token 309 Class A shares for about $226 million; the second quarter brought $4.5 billion, and July added more than $3.3 billion. Abel himself bought 21 Class A shares for roughly $14.6 million, approximately the after-tax value of his annual salary.
Taken together that is a more coherent statement than any single portfolio line: the company considers its existing holdings good enough to leave alone, finds exactly one outside opportunity at the required scale, and spends the rest buying itself.
The threshold nobody ever indexed
For a U.S. investor there is a wrinkle worth knowing, because it determines how much of the market you actually get to see. The $100 million reporting threshold that decides who files a 13F was set in the late 1970s and has never been adjusted for inflation. In dollars of the era it was a genuinely large institution; today it captures thousands of managers, which is why the filings you read include firms you have never heard of. The SEC proposed raising it to $3.5 billion in 2020, drew heavy opposition from issuers who use the data to identify their own shareholders, and withdrew the proposal in 2021. The threshold that governs the entire ritual is, in real terms, a small fraction of what Congress originally had in mind.
It is also worth knowing which American disclosure regimes are actually fast, because 13F is the slow one. A Schedule 13D — the filing that signals intent to influence a company — is due within days of crossing five percent, not six weeks after quarter end, and it requires the filer to state the purpose of the acquisition. Passive holders above five percent file the lighter 13G. Corporate insiders file a Form 4 within two business days of a trade. If you want to know what somebody intends rather than what they owned six weeks ago, 13F is the wrong document. It is simply the most famous one.
Finally, the access point in this story was closed to individuals from the start. The Alphabet placement at $351.81 and $348.20 was a negotiated transaction between an issuer that needed $80 billion in a hurry and a buyer able to write a $10 billion check without selling anything first. No brokerage account gave you that price. The retail equivalent of Berkshire’s quarter was available all along in the open market — which is precisely where the other 19.57 million shares came from.
What argues against this reading
Criticism of the 13F can be overdone, and routinely is. Three objections deserve serious weight.
First, the form is a much better indicator for an investor like Berkshire than for an active hedge fund. When 72 percent of a portfolio is unchanged for an entire quarter, a six-week delay is close to irrelevant. The staleness critique lands on turnover, not on buy-and-hold.
Second, there is genuine empirical evidence that copying can work. A widely cited 2013 paper by Aiken and coauthors finds that portfolios replicating the holdings of consistently successful managers still generate excess returns even after the 45-day lag. Anyone claiming 13Fs are worthless is arguing against the data.
Third, this filing was more transparent than usual, not less: no positions were withheld confidentially, and the largest single item had already been announced in a June 1 press release. The form’s marginal contribution here was mainly to make the unannounced portion of the Alphabet purchase visible. That is not a small contribution.
The defensible core of the critique is narrower and survives: the form is good for reconstructing what a manager owns. It is not good for inferring what they think, and it is worse still for determining what a differently capitalized investor ought to do.
Four questions for any institutional filing
This quarter yields a short checklist that will work in every future filing season.
First: reconcile by CUSIP, never by name. Bank of America and Chubb between them would have faked four transactions worth $78 billion in this filing, not one of which occurred.
Second: check whether the change is already explained by a disclosed event. About 60 percent of the Alphabet increase came from a placement announced on June 1. The interesting part is the remainder.
Third: ask what constraint the filer is operating under. A firm with $344 billion of liquidity and a concentrated fund with a dozen positions can trade the same stock in opposite directions without disagreeing about it at all.
Fourth: ask what is structurally invisible in this reporting class. For Berkshire that means the Japanese trading houses, the bond portfolio and the wholly owned subsidiaries — which is to say, most of the company.
What remains is the finding the filing season opened with. The most closely watched portfolio document of the quarter contained a single new position worth about as much as a used sedan, and alongside it, spread unobtrusively across two lines, the second-largest equity purchase this company has made in years. Both were in the same document, in the same place, legible to anyone. The difference was not access to the information. It was whether anybody did the arithmetic.
Try TradingView Free for 30 Days
Plus get a $15 discount on your first subscription through this link.
