Apple's diluted share count was 188 million above the shares it had at year-end. Basic, diluted and the cover page: three share counts, and which one to use.
Every figure below was read from the companies' own filings on sec.gov and is cited with its accession: Apple's FY2026 10-K, Alphabet's 10-Q for the quarter to 30 September 2026 and SpaceX's 10-Q for the same quarter. We show no share prices, no market capitalisations, no analyst estimates and no consensus. Differences, percentages, vote shares and per-share figures the companies do not print are our arithmetic on filed values and are marked (Ours.) Research, not investment advice.
Apple's 10-K for the fiscal year to 26 September 2026 (0000320193-26-000026) prints three share
counts, and none of them matches another:
| Apple, FY2026 | Shares | Where it is |
|---|---|---|
| Outstanding on 23 October 2026 | 14,532,182,000 | cover page |
| Outstanding on 26 September 2026 | 14,532,259 thousand | balance sheet, Note 11 |
| Weighted-average basic, FY2026 | 14,661,935 thousand | Note 3 |
| Weighted-average diluted, FY2026 | 14,720,306 thousand | Note 3 |
(Ours: the diluted count is 188.0 million shares, or 1.3%, above the shares outstanding at year-end; the basic count is 129.7 million above it.) Diluted earnings per share were $8.98. Divide the same $132,175 million of net income by the year-end count and you get about $9.10 (Ours; not a filed figure).
Neither number is wrong. They answer different questions, and this post is about which count answers which.
What are the three share counts?
Shares outstanding is a count on a date. The cover page of a 10-Q or 10-K gives it as of a day shortly before filing, after the period ends. The balance sheet gives it as of the period end. It is the number of shares that exist and are held outside the company on that day.
Weighted-average basic shares is an average over the period, weighted by how many days each share was outstanding. A share bought back halfway through the year counts for half a year. It is the denominator of basic earnings per share, because earnings also accrue over the period.
Weighted-average diluted shares adds the shares that would exist if dilutive instruments (restricted stock units, options, convertible securities) turned into stock, and only when adding them lowers earnings per share. It is the denominator of diluted EPS.
Two of the three are averages built for an earnings ratio; one is a count on a date.
Which count goes into market cap?
Shares outstanding, on the same date as the price you multiply by. Market cap is a value on a day, so it needs a count on a day. The cover page is the most recent count the company files; the balance sheet count is the period-end alternative.
Apple's cover: "14,532,182,000 shares of common stock were issued
and outstanding as of October 23, 2026" (0000320193-26-000026). Against the balance sheet's
14,532,259 thousand at 26 September, that is about 77,000 fewer shares four weeks later (Ours).
With several classes you add every class, listed or not, because each has the same economic claim.
Why is Apple's diluted count above the shares it actually has?
Two reasons, both in the 10-K.
The buyback runs ahead of the average. Note 11 reconciles the year (0000320193-26-000026),
in thousands:
| Apple, FY2026 | Shares (thousands) |
|---|---|
| Common stock outstanding, beginning balance | 14,773,260 |
| Common stock repurchased | (296,699) |
| Common stock issued, net of shares withheld for employee taxes | 55,698 |
| Common stock outstanding, ending balance | 14,532,259 |
The 10-K's sentence: "During 2026, the Company repurchased 297 million shares of its common stock for $85.8 billion." A company that shrinks its count all year has a weighted average above its year-end count, because the early months still carry the shares it later retired. (Ours: a net reduction of 241.0 million shares over the year.)
Dilution is counted with the treasury stock method. Apple had 142,405 thousand RSUs outstanding at 26 September, and $24.1 billion of unrecognized compensation cost on them, to be recognized "over a weighted-average period of 2.6 years". Only 58,371 thousand shares enter the diluted count as the "Effect of dilutive share-based awards". The treasury stock method assumes the unrecognized cost buys back part of the shares, so the add-on is smaller than the award count.
For per-share value at a date, the year-end count plus the awards you expect to settle in stock measures the claim on the company more directly than a past year's average.
How do you count a company with three classes?
Alphabet has Class A (one listed line), Class C (another listed line) and Class B, which is not
listed. Its 10-Q for the quarter to 30 September 2026 (0001652044-26-000084) states on the cover:
"As of October 21, 2026, there were 5,888 million shares of Alphabet's Class A stock outstanding,
827 million shares of Alphabet's Class B stock outstanding, and 5,579 million shares of Alphabet's
Class C stock outstanding."
(Ours: 5,888 + 827 + 5,579 = 12,294 million shares.) The balance sheet gives the same total at 30 September, 12,294 million, with a different mix: Class A 5,888, Class B 828, Class C 5,578. A market cap built from one listed line alone would leave out more than half of the shares.
The 10-Q explains why the classes can be added: the rights of Class A, B and C, "including the liquidation and dividend rights", "are identical, except with respect to voting". So Alphabet computes EPS with the two-class method and gets the same figure for each class, give or take a cent. Note 12 for the third quarter, in millions of shares:
| Alphabet, Q3 2026 | Class A | Class B | Class C | Consolidated |
|---|---|---|---|---|
| Basic weighted average | 5,878 | 830 | 5,548 | 12,256 |
| Add: conversion of Class B to Class A | 830 | — | — | — |
| Add: RSUs and other contingently issuable shares | — | — | 112 | 112 |
| Add: conversion of preferred stock | 0 | 0 | 0 | 0 |
| Diluted weighted average | 6,708 | 830 | 5,660 | 12,368 |
| Diluted EPS | $2.12 | $2.12 | $2.11 | $2.12 |
Two details. The Class A diluted column already includes the 830 million Class B shares, because "The computation of the diluted net income per common share of Class A stock assumes the conversion of Class B stock"; add the Class A diluted column to the Class B column and you count those shares twice. Use the consolidated column. And the RSU add-on of 112 million sits under Class C. Against the 284 million unvested RSUs and PSUs the 10-Q lists at 30 September, it is the same treasury stock effect as at Apple. (Ours: the consolidated diluted count is 112 million above basic, and 74 million above the cover count.)
What about convertible preferred and an at-the-market program?
Alphabet added two share sources in June that a cover count does not show.
Mandatory convertible preferred. On 5 June the company issued "385 million Series A and Series B
depositary shares, representing 19 million shares of 6.25% Mandatory Convertible Preferred Stock",
for net proceeds of $19.0 billion. Each preferred share converts on or about 15 May 2029 into
"between 2.2520 and 2.8160 shares of Class A stock" (Series A) or "between 2.2740 and 2.8420 shares
of Class C stock" (Series B), depending on the stock's value at conversion (0001652044-26-000084).
(Ours: 385 million depositary shares at 1/20 each is about 19.25 million preferred shares, half per series. At the minimum rates that is about 43.6 million new common shares; at the maximum, about 54.5 million.) The company also bought capped calls for $1.0 billion that are "intended to reduce the potential dilution", so the net share effect at conversion can be lower.
In the diluted count the preferred appears only when it is dilutive. In the third quarter it added zero shares. For nine months it added 24 million, and the diluted count was 12,324 million against 12,169 million basic. The footnote: preferred shares "are assumed to have been converted to common shares only if they have a dilutive effect on net income per common share". A quarter's diluted count can leave out a security that will convert by contract.
The ATM program. Alphabet may issue up to $40.0 billion of Class A and Class C stock "through an at-the-market offering program". In the three months to 30 September it "issued 12 million shares of Class A stock for net proceeds of $4.3 billion and 24 million shares of Class C stock for net proceeds of $8.1 billion", and $27.6 billion remained available. Repurchases in the quarter were zero. (Ours: 36 million new shares for $12.4 billion in one quarter.) Shares sold late in the quarter are in the cover count in full but in the weighted average only for the days they existed.
When the cover is far above the average: SpaceX
SpaceX's 10-Q for the quarter to 30 September (0001628280-26-066880) shows how far the counts
drift apart in a year of corporate events. The cover: "As of October 15, 2026, the registrant had
8,246,541,218 shares of Class A common stock and 5,334,522,491 shares of Class B common stock
outstanding." Note 13 for the quarter:
| SpaceX, Q3 2026 (millions) | Shares |
|---|---|
| Cover count, 15 October (Class A + Class B) (Ours) | 13,581.1 |
| Balance sheet, 30 September (8,240 + 5,339) (Ours) | 13,579 |
| Weighted-average basic, Q3 | 11,942 |
| Add: stock options | 457 |
| Add: restricted stock units | 75 |
| Add: ESPP | 1 |
| Weighted-average diluted, Q3 | 12,475 |
| Weighted-average basic and diluted, nine months | 7,274 |
(Ours: the cover count is 1,639 million shares above the quarter's basic average.) The 10-Q gives the events: the IPO in June, in which SpaceX "issued and sold 638.9 million shares of its Class A common stock"; the conversion at the IPO of "all outstanding shares of the Company's redeemable convertible preferred stock" into Class A and Class B; a five-for-one stock split in May; and the Cursor merger, completed on 14 August, paid in Class A shares. Shares issued in mid-August count for only part of the quarter. For the nine months, with a net loss, basic and diluted are the same 7,274 million: adding shares to a loss would make the loss per share smaller, so nothing is added.
Three more lines a cover count does not carry:
- Excluded awards. The diluted count leaves out "1,333 million" share-based awards "subject to performance and market conditions that were not met" at 30 September, plus 19 million anti-dilutive awards.
- Contracted shares. The EchoStar spectrum purchase includes "approximately $11.1 billion in equity, payable through the issuance of approximately 261.8 million shares" of Class A, which "will be issued at the Spectrum Acquisition Closing". They are in no count yet.
- Votes are not shares. "Each holder of Class A is entitled to one vote for each share held. Each holder of Class B is entitled to ten votes for each share held." (Ours: on the cover counts, Class B is 39.3% of the shares and 86.6% of the votes.) For value, every share counts once; for control, the classes count very differently. Class B "is convertible at the option of the holder, at any time, into one share of Class A", and converts automatically on most transfers, which is why Class B fell from 5,339 million at 30 September to 5,334.5 million on 15 October while Class A rose.
Which count do I use for per-share value?
A short rule set, each line traceable to a filing:
- Market cap: shares outstanding on the cover page, all classes added, at the same date as the price. Never one listed class alone.
- Per-share intrinsic value: start from the most recent outstanding count, then add what will become stock: RSUs and options you expect to settle in shares, convertibles at the conversion terms, contracted issuances. Write each addition down with its accession.
- Earnings per share: use the company's own diluted EPS and diluted weighted average together. Do not mix a period's earnings with a later date's count unless you say so.
- Multi-class companies: use the consolidated column. A class column that already assumes conversion double-counts if you add the other class to it.
- Losses: in a loss period diluted equals basic by rule; read the excluded lines in the EPS note.
Our buybacks-to-bonds post follows the repurchase line across the hyperscalers; this one is the counting mechanic underneath it.
Run it yourself
Open /filings/AAPL, take the FY2026 10-K, and read three places: the cover page, Note 3 and Note 11. Then do the same for /filings/GOOGL with the third-quarter 10-Q: the cover page, Note 11's preferred stock and ATM paragraphs, and Note 12.
In the DCF calculator, set the share count yourself, and write down which of the three counts you used and what you added to it. Our Thesis Monitor reads each new filing against what you wrote, so the cover count and the diluted add-ons are lines worth putting in the thesis. If you work inside an agent, the same filings are on the EvidInvest MCP server. It is free to start, credit packs from $10, and there is no subscription.
The cover page tells you how many shares exist on a day. The EPS note tells you how many existed on average, and the equity note tells you how many more are already promised.
Sources. Apple Inc. FY2026 Form 10-K for the fiscal year ended 26 September 2026, filed 3 November
2026: accession 0000320193-26-000026
(sec.gov).
Alphabet Inc. Form 10-Q for the quarter ended 30 September 2026, filed 29 October 2026:
0001652044-26-000084
(sec.gov).
Space Exploration Technologies Corp. Form 10-Q for the quarter ended 30 September 2026, filed
22 October 2026: 0001628280-26-066880
(sec.gov).
EDGAR was checked at 04:30 UTC on 8 November 2026; no newer 10-Q or 10-K had been filed by Apple,
Alphabet or SpaceX. No share price, market capitalisation, analyst estimate, consensus figure or
call remark appears on this page. Differences, percentages, vote shares, conversion ranges in shares
and per-share figures are arithmetic on those filed values, labelled as ours where they appear.
Research, not investment advice.
Working with this data from an AI agent? The EvidInvest MCP server gives Claude, Cursor, and any MCP client access to 46 financial data, valuation, and SEC intelligence tools.
Fair Value Weekly
Get DCF breakdowns, fair value updates, and portfolio ideas for serious investors. No spam, no paywalled teasers.