The Slice a Single Share Represents
EPS = net income ÷ weighted-average shares outstanding
It answers one narrow question: of the accounting profit recorded this period, how much is attributable to each share. It is not a measure of the company’s size, its cash generation, or its value.
The denominator is a weighted average rather than a closing count for a practical reason. The share count moves all year: buybacks retire stock, equity compensation issues it. Measuring twelve months of profit against a single day’s count would misstate the result. The weighted average reflects how many shares existed for how much of the period.
Running the Formula on Apple’s Filing
Apple’s fiscal year 2025 ended September 27, 2025. From the statements filed with its Q4 FY2025 Form 8-K:
Check it yourself: $112,010M ÷ 14,948.5M shares = $7.4931, rounding to the $7.49 reported. The formula hides nothing. The interpretation is where the work is.
Basic and Diluted: A 56-Million-Share Gap
Companies report EPS twice. Basic uses shares actually outstanding. Diluted adds shares that would come into existence if every in-the-money option, restricted stock unit and convertible were exercised.
Apple’s fiscal 2025 basic count was 14,948.5 million against a diluted count of 15,004.7 million, a difference of 56.2 million shares, roughly 0.4%. Those are equity awards not yet converted to ordinary stock, and they move basic EPS of $7.49 down to diluted EPS of $7.46.
Diluted is the figure to track, because those shares are a real claim on future profit. At Apple the gap is negligible. At a company paying much of its workforce in equity the gap runs to several percent, and the basic figure flatters the result materially.
When a headline quotes EPS without saying which version, assume it is whichever number looks better. Both are disclosed on the income statement of every 10-Q and 10-K on EDGAR, with the share counts used. Checking takes under a minute.
Decomposing Apple’s 22.7%
Diluted EPS went from $6.08 to $7.46, or +22.7%. Three separate things produced that, and only one of them is Apple selling more.
A starting point that was artificially low
On September 10, 2024, the European Court of Justice reinstated the European Commission’s State Aid decision, and Apple recorded a one-time income tax charge of $10.2 billion net in fiscal 2024: $15.8 billion payable to Ireland, offset by a $4.8 billion US foreign tax credit and an $823 million reduction in unrecognized tax benefits. In the same filing, Apple published adjusted fiscal 2024 figures excluding it: net income of $103,982 million and diluted EPS of $6.75.
Measured against $6.75 rather than $6.08, fiscal 2025 diluted EPS grew 10.5%, not 22.7%. Both numbers are arithmetically correct. They describe different things.
A denominator that shrank
Basic shares dropped 2.6%, from 15,343.8 million to 14,948.5 million — roughly 395 million shares retired through repurchases. The same profit across fewer shares produces a bigger number per share, worth about 3 percentage points of the GAAP growth rate and 2.8 points of the adjusted one.
And, third, the business itself
Revenue rose 6.4%. Adjusted net income rose 7.7%. That is a good year for a company of Apple’s size, and it is not 22.7%.
A 22.7% headline that is really about 7.7% of business growth, plus a distorted base, plus a buyback, is not a scandal. It is ordinary financial reporting, and it is exactly why the headline alone is not enough.
The Buyback Effect Is Market-Wide
Apple is not unusual here. S&P 500 companies repurchased a record $293.5 billion of their own stock in the first quarter of 2025 alone, and $999.2 billion over the preceding twelve months, with 384 index members repurchasing at least $5 million each (S&P Dow Jones Indices, 2025). Across the index, a meaningful share of reported EPS growth in any period is denominator arithmetic.
The check that takes thirty seconds: put EPS growth and net income growth side by side. If EPS is growing meaningfully faster, look at the share count. If EPS is growing while net income is flat or falling, the growth is entirely buyback-driven and the business is not improving.
GAAP EPS and the Company’s Own Version
Alongside the audited GAAP figure, most companies publish an adjusted EPS with items management regards as unrepresentative removed: restructuring, acquisition costs, litigation, and frequently stock-based compensation.
Apple’s tax-charge adjustment is the honest version of this: a genuinely singular event, disclosed with a full reconciliation, and the adjusted figure produced the lower growth rate rather than the flattering one. Treat that as the benchmark for good practice.
The pattern worth distrusting is the opposite. Three things are worth checking in any reconciliation table:
- Exclusions that recur. Restructuring charges in eight consecutive quarters are not one-time by any ordinary meaning of the word; they are the cost of running that business.
- Stock-based compensation removed. The single most consequential adjustment in technology. It is a real cost, settled in shares that dilute the holders reading the adjusted figure.
- An adjusted figure with no GAAP equivalent nearby. The reconciliation must be presented; a release that buries it pages from the headline is signalling which number it wants quoted.
A quick sanity test on any adjusted EPS: add up the excluded items across the past three years. If the total is a large fraction of cumulative GAAP profit, those “adjustments” are not incidental to the business — they are a substantial part of it.
Why Two Companies’ EPS Cannot Be Compared
This is the most common misuse of the figure. EPS depends on how many shares a company happens to have issued, and that number is a historical accident of financing decisions rather than a fact about the business.
Consider two companies earning identical profit:
Company A’s EPS is five times Company B’s. The businesses are identical, valued identically, and an investor putting $10,000 into either owns exactly the same claim on $500 million of profit. The only difference is that one company divided its equity into more pieces.
This is also why a stock split changes EPS and changes nothing else. A two-for-one split halves EPS overnight because the share count doubles; no shareholder is worse off. Any comparison that treats a higher EPS as a better company is reading the share count, not the performance.
Trailing, Forward, and the Estimate EPS Is Judged Against
Three versions circulate, and they answer different questions. Trailing EPS is the last four reported quarters: actual, audited, backward-looking. Forward EPS is the analyst consensus for the next four quarters — an estimate, and therefore an opinion. Current-quarter consensus is the specific number a company is measured against on earnings day.
That last one shapes how stocks move, and the scoring deserves a look. In the second quarter of 2026, 86% of S&P 500 companies reported EPS above estimates. The five-year norm is 78%, the ten-year 76% (FactSet, 2026).
Roughly three companies in four beat expectations, every quarter, for a decade. That is not evidence that management consistently outperforms — it is evidence that the bar is set where it can be cleared. Companies guide analysts toward achievable numbers in the weeks before reporting, so a “beat” is closer to the base case than to good news. A stock can beat consensus and fall the same morning, because the market was pricing something better than the published estimate.
The magnitude data in the same report also shows how badly one item can distort an aggregate. Companies reported earnings 39.3% above estimates in that quarter, against a five-year norm of 7.0%. Strip out a single $98 billion gain at Alphabet and the figure was 12.6% (FactSet, 2026). One company moved the whole index statistic by 27 percentage points. It is the same lesson as Apple’s tax charge, at market scale.
Five Things the Number Cannot Tell You
- It is not comparable across companies. EPS depends on how many shares a company happens to have issued. Two identical businesses with different share counts report different EPS. This is why the P/E ratio exists.
- It is accounting profit, not cash. Net income includes non-cash items. A company can report healthy EPS while free cash flow deteriorates.
- It can be engineered upward. Buybacks raise EPS whether or not the business improved.
- One period tells you little. Seasonality, one-time items and base effects all distort a single comparison, as Apple’s fiscal 2024 shows.
- Negative EPS is not disqualifying. A company investing ahead of revenue may post losses by design, though the case then rests entirely on future earnings.
How EPS Feeds the P/E Ratio
EPS matters most as the denominator of price-to-earnings: share price ÷ EPS. Because P/E inherits EPS, it inherits every problem above. A P/E built on adjusted EPS is not comparable with one built on GAAP EPS.
The distortion can become total. The S&P 500’s trailing P/E reached 123.73 in May 2009 — not because stocks were expensive, but because aggregate earnings had collapsed, leaving almost nothing in the denominator (multpl). When EPS approaches zero, every ratio built on it stops working.
The Bottom Line
EPS is standardized, audited, and narrower than the weight placed on it. It tracks one company against its own past well and compares two companies badly. It reflects accounting profit rather than cash. And it moves on share count and base effects as readily as on performance.
Read the diluted figure. Keep GAAP and adjusted separate. Put EPS growth next to net income growth every time, and check what the prior-year base contained. Apple’s 22.7% was three different stories stacked on top of each other — and the one about the business was the smallest of them.
For what sits underneath EPS, see net income explained and outstanding shares.
