How to Adjust Per-Share Value for Heavy Buyback Activity

AutoZone grew diluted earnings per share 13.0% in fiscal 2024, from $132.36 to $149.55. Net income grew 5.3%. The gap between those two numbers — worth nearly eight percentage points — didn't come from margin expansion or a hot quarter of same-store sales. It came from the denominator. AutoZone's own fiscal 2024 10-K attributes $0.96 of that year's EPS increase specifically to a shrinking share count, built from $3.2 billion spent buying back 1.1 million shares at an average price of $2,759.

That's the part of per-share valuation most models handle sloppily. An owner earnings estimate, a DCF, a fair value multiple — all of them typically start with a company-wide number and divide by today's share count to get a per-share figure. Fine for a single point in time. But it breaks down the moment you project forward, because the share count a year or five years from now isn't fixed. It's the product of capital allocation decisions a company is actively making right now, and those decisions can move the denominator almost as much as operations move the numerator.

Two companies make the point better than any formula on its own. One has spent a quarter-century treating its share count as the primary lever of shareholder value. The other spends real money on repurchases every year and its share count keeps rising anyway. Same word — buyback — doing opposite things to per-share value, depending on what's happening on the other side of the ledger.

Owner Earnings Divided by a Number That Isn't Fixed

Per-share value starts from a simple identity: take a company's total owner earnings — net income plus depreciation and amortization, minus the capital expenditures required to keep the business running — and divide by diluted shares outstanding. Most valuation work treats the share count in that equation as background noise: whatever the filing says this quarter, carried forward with a small annual drift assumption tacked on almost as an afterthought.

That treatment is defensible for a company issuing a percent or two of new shares a year through ordinary employee compensation, quietly offset by a token buyback program. It stops being defensible the moment a company treats its share count as a genuine capital allocation lever — spending double-digit percentages of market cap on repurchases in some years, or growing its share base through stock-based compensation faster than any buyback program can absorb. For those businesses, projecting per-share value forward means projecting two growth rates, not one: the growth rate of aggregate owner earnings, and the growth — or shrinkage — rate of the share count dividing it.

Per-Share Value Growth ≈ Growth in Total Owner Earnings − Growth in Diluted Share Count (A first-order approximation. The exact relationship is multiplicative: Per-Share Growth = [(1 + Owner Earnings Growth) ÷ (1 + Share Count Growth)] − 1. The approximation holds reasonably well under roughly 15% growth rates in either variable; beyond that, use the exact form.)

Name each side of that equation separately before combining them. A business compounding owner earnings at 8% a year with a flat share count delivers 8% per-share growth. The same business shrinking its share count 3% a year through buybacks delivers something closer to 11%. A business compounding owner earnings at 8% while its share count grows 3% a year through equity issuance delivers something closer to 5%. Same underlying business, same operating performance. A four-to-six-point swing in per-share value growth, purely from what management does with the denominator.

AutoZone: The Denominator as the Primary Engine

AutoZone has never paid a common dividend. Since 1998 it has instead directed essentially all of its excess capital toward its own stock, repurchasing roughly 155 million shares for around $38 billion through May 2025 — cutting its share count by close to 90%, from the neighborhood of 150 million shares outstanding in the late 1990s to under 18 million today.

Fiscal 2024 (year ended August 31, 2024) shows the mechanism in a single year's numbers. Net income came in at $2,662.4 million, up 5.3% year over year. Diluted EPS rose 13.0%, from $132.36 to $149.55. The company repurchased 1.1 million shares during the year for $3.2 billion, and its own filing states that the resulting share-count reduction added $0.96 to diluted EPS relative to what it would otherwise have been.

Run the formula above on those same numbers and the mechanism is visible directly, even without the company's own disclosure. Net income growing 5.3% with a flat share count would have put diluted EPS at roughly $139 — 5.3% above the prior year's $132.36. Actual EPS came in at $149.55. Most of the remaining gap traces to the shrinking denominator, broadly consistent with the $0.96 buyback-specific impact the company discloses directly. I'd treat any attempt to split that residual precisely between "this year's repurchases" and "the compounding effect of a share base built down over 25 years" as a directional exercise rather than an exact one — the two effects blend together in a business that's been running this program for a quarter-century, and untangling them cleanly from one year's filing alone isn't really possible.

The trade-off shows up elsewhere on the balance sheet. AutoZone ended fiscal 2024 with total stockholders' equity of roughly negative $4.75 billion — a quarter-century of buybacks and retained deficits running past retained earnings — funded partly by net debt near $12 billion. That's not automatically a distress signal for this particular business; it's a capital structure choice a management team with a long, consistently disclosed track record has made deliberately, one that would look reckless at a company with less durable or less predictable cash generation. The point for a per-share valuation exercise isn't whether the debt load is prudent — that's its own separate question — it's that AutoZone's per-share value has been engineered, disclosed year after year, in a way owner earnings alone won't show you if you only look at the numerator.

Snowflake: When the Denominator Moves the Wrong Way

Not every buyback program shrinks the share count, and conflating "the company is repurchasing stock" with "the share count is falling" is one of the more common mistakes in per-share valuation work. Snowflake spent roughly $1.9 billion repurchasing about 14.8 million shares in fiscal 2025 (year ended January 31, 2025) — real money, a real program, reported the same way any buyback is reported.

And diluted shares outstanding rose anyway — from a weighted average of roughly 328.0 million in fiscal 2024 to about 332.7 million in fiscal 2025, an increase of nearly 1.4%. The reason is straightforward: stock-based compensation ran to $1.479 billion for the year, up from $1.168 billion the year before, and new shares issued to employees outpaced what the buyback program bought back. The repurchases were real. Money spent, honestly reported. They just weren't large enough to offset the new shares being created on the other side of the ledger.

This is the case a per-share valuation model has to catch, and a lot of them don't. A screener or a quick model that sees "$1.9 billion in buybacks" and assumes the share count fell has the sign backwards. For a business with heavy stock-based compensation — most growth-stage software companies fall into this category — the buyback program's real job, in most years, isn't shrinking the share count at all. It's slowing the rate at which the share count grows. Those are different jobs, and only one of them adds per-share value the way AutoZone's program does.

The Discipline Rule Most Per-Share Models Skip

None of this makes buybacks automatically good for per-share value, even in years when they do shrink the share count. A repurchase only creates value for continuing shareholders when it's executed at a price below the business's intrinsic value — buy back stock above what it's actually worth and the transaction transfers value from the shareholders who stay to the ones who sell, dressed up in a mechanism that looks identical on the cash flow statement either way.

That discipline gets harder to hold in some rate environments than others. As Thomas Brennan has covered in how Fed policy regimes shape sector leadership, a rising-rate regime raises the discount rate applied to a company's own future cash flows the same way it raises the discount rate an outside investor would apply to them — which means the bar an internal buyback has to clear to be accretive moves with the cycle too, even though the capital allocation decision is being made by management rather than by the market pricing the stock day to day.

The honest version of a buyback-adjusted per-share model carries this as an explicit check, not an assumption. Project the share count forward using the company's stated repurchase authorization and its historical stock-based compensation issuance rate — netted against each other, not just the buyback side of the ledger — and separately ask whether the price the company is likely paying sits below your own fair value estimate. A business can pass the first test and fail the second. So a per-share value estimate that only checks the first is optimistic in a way that won't survive management paying up for its own stock at the wrong point in the cycle. A buyback-adjusted per-share estimate is still a compass, not a GPS coordinate — the share-count projection just keeps the compass from pointing somewhere the denominator has already ruled out.

Key Takeaways

• Per-share value depends on two growth rates, not one: the growth of aggregate owner earnings and the growth — or shrinkage — of diluted shares outstanding. Treating the share count as fixed is a reasonable simplification only when it's genuinely stable.
• AutoZone's fiscal 2024 10-K shows the mechanism directly: 5.3% net income growth became 13.0% diluted EPS growth, with the company itself attributing $0.96 of the increase to a shrinking share count built from roughly $38 billion in cumulative repurchases since 1998.
• Snowflake's fiscal 2025 numbers show the opposite: a real $1.9 billion buyback program that still left diluted shares outstanding higher than the year before, because stock-based compensation issuance outpaced it.
• A buyback only adds per-share value when it's executed below intrinsic value. Projecting the share count forward is a mechanical exercise; judging whether management is paying a smart price for its own stock is a separate, harder question that per-share models routinely skip.