Capital Allocation

For its first decade as a public company Salesforce bought growth: roughly $50 billion of goodwill from Tableau, Slack, MuleSoft and Informatica. Through FY2022 that capital earned close to nothing — about a 1% return on invested capital. The FY2024 margin reset lifted the figure to near 10%, roughly its cost of capital, and management turned the cash machine toward buybacks and a dividend. FY2026 did both at once, funded by $25 billion of new debt.

Two machines

Salesforce has run two capital-allocation regimes back to back. From FY2017 to FY2022 it was a serial acquirer, converting cash and its own shares into scale. Goodwill on the balance sheet grew from $7.3 billion to $57.9 billion — a $50.6 billion pile of premium paid over the net tangible assets of the companies it absorbed [1]. Four deals account for most of it.

Source: Salesforce balance sheets, FY2017–FY2026; step-ups from MuleSoft (FY2019), Tableau (FY2020), Slack (FY2022) and Informatica (FY2026) [2].

The headline transactions set the scale of the ambition. Tableau closed in 2019 for $15.7 billion, paid entirely in stock, and brought $10.9 billion of goodwill [3]. Slack closed in July 2021 for $27.1 billion — $15.8 billion in cash and $11.1 billion in stock — the largest deal in the company's history and $21.4 billion of it goodwill [4]. Informatica followed in FY2026 at $9.6 billion, almost all cash [5].

Sources: Tableau prospectus [6]; Vlocity and Acumen [7]; Slack [8]; Regrello [9] and Informatica [10].

The $85 billion figure often attached to Salesforce's dealmaking is a lifetime headline that also captures earlier purchases — ExactTarget, Demandware, MuleSoft. The cleaner measure of what the buying cost is the goodwill still carried today: $57.9 billion, or 52% of total assets, none of it deductible for tax and none of it ever written down. That last point cuts two ways, and the return math is where it resolves.

What the acquisitions earned

The honest test for a serial acquirer is return on invested capital measured with goodwill in the base — because the goodwill is the price paid, and a buyer only creates value if the acquired earnings clear the cost of the capital sunk into them. On that basis the picture is a straight line up.

Source: derived from Salesforce consolidated financials, FY2022–FY2026 — NOPAT (operating income less cash taxes) over stockholders' equity plus debt less cash; FY2026 operating income $8.3B [11], invested capital ~$66B [12].

Through FY2022 the returns were roughly 1%. On the capital the market could see, the acquisitions had built a much larger company that generated almost no economic profit — the criticism that drew activist investors to the register in early 2023. What changed was not the deals but the cost base: FY2026 operating income of $8.3 billion is fifteen times the $0.55 billion of FY2022 [13], while invested capital barely moved. That lifted goodwill-inclusive ROIC to about 10% — close to the 10% discount rate the reverse-DCF in What the Price Implies applies.

The read this supports: the acquisitions now roughly earn their keep, but only just, and only because of the margin reset. Two facts sit against a stronger conclusion. First, the return sits at the cost of capital, not comfortably above it — collectively the $50 billion of goodwill has yet to prove it compounds value rather than merely preserving it. Second, the calculation is conservative in Salesforce's favour: it counts only cash as excess, so including the marketable-securities balance would shrink invested capital and lift the return by a point or two. The absence of any goodwill impairment across a decade is consistent with acquisitions that pull their weight; it is also what a company under no accounting pressure to admit an overpayment would report. The read is most sensitive to whether the margin reset holds, which connects directly to the growth question in Growth and Agentforce: the returns stay near 10% only while the operating leverage of the last three years persists.

The pivot to returns

The clearest signal of how management now reads its own opportunity set is where the cash goes. Having spent a decade buying companies, Salesforce began in FY2023 to buy its own stock, and in FY2025 initiated a dividend.

Source: Salesforce consolidated statements of cash flows, FY2023–FY2026, as reported.

Across FY2023–FY2026 the company returned about $35 billion — roughly $32 billion of buybacks and $3 billion of dividends — against $42.6 billion of cumulative free cash flow, or about 83% of what it generated. The buyback authorization was lifted repeatedly, from $10 billion at inception in August 2022 to a $50 billion total by February 2026 [14]. The dividend, first paid in April 2024 at $0.40 a quarter, was raised to $0.416 through FY2026 and then to $0.44 [15]. For an investor who treats buybacks as evidence of discipline, the size is settled; the read is most sensitive to the price paid.

On price, the record is mixed. The repurchases have been programmatic rather than opportunistic, and much of the earlier volume was bought at prices the stock has since fallen below.

Sources: FY2026 Q1 open-market at $273.42, ASR initial delivery of 103M shares at $198.34, and FY2027 Q1 open-market at $192.00 [16]; FY2026 average $254.21 as established in The Cash Machine; recent quote ~$171 per What the Price Implies.

Every tranche shown was bought above the recent ~$171 quote, including the cheapest. That is the fair charge against the buyback: it has shrunk the share count but has not, so far, been timed to the lows. The countervailing fact is the March 2026 accelerated share repurchase. Salesforce committed $25 billion — half the entire authorization — in a single stroke, taking initial delivery of about 103 million shares at $198.34, funded not from cash but from $25 billion of new senior notes maturing between 2028 and 2066 [17], [18]. Coming after the stock had de-rated by more than half from its December 2024 peak, this was a leveraged bet on the company's own undervaluation — the balance-sheet transformation examined in Cash Quality, read here as a capital-allocation choice. It moved Salesforce from a near-net-cash position to roughly 2x net debt to free cash flow. How that judgment reads is most sensitive to the stock price, which has since fallen a further 14% below the ASR's own average price.

Informatica and the discipline question

The tidy story — acquirer becomes disciplined capital returner — is complicated by FY2026, when Salesforce did everything at once. Alongside the record $12.6 billion of buybacks and the first dividend increase, it closed the $9.6 billion Informatica purchase and the smaller Regrello deal, and it did so on borrowed money: a $4 billion and a $2 billion credit facility taken in June 2025 to finance Informatica, later refinanced into a $6 billion five-year term loan [19].

Informatica is the clearest read on whether the post-2023 restraint is a permanent change in temperament or a pause. It is the first large, debt-financed acquisition since Slack, struck by a management team that spent two years telling investors it had learned to prize profitability over scale. The deal is defensible on strategy — data management feeds directly into the Data 360 franchise that anchors the AI case in Growth and Agentforce — but it lands on the same balance sheet already carrying $50 billion of not-yet-proven goodwill, and it was funded with debt at a moment when the company was also borrowing $25 billion to buy back stock. The reader learns more about management's true priorities from watching whether Informatica is the last big deal or the first of a new wave than from any statement on a call.

What would change the read

Three things would move this assessment. Goodwill-inclusive ROIC drifting back above 11–12% would confirm the acquisitions compound value rather than merely cover their cost; a slip back toward the mid-single digits, most likely through margin give-back, would revive the pre-2023 verdict that the buying built size without economic profit. A goodwill impairment — absent for a decade — would be the first accounting admission that a price was too high. And a second Informatica-scale, debt-funded deal within the next year or two would signal that the discipline of FY2023–FY2025 was cyclical, not structural. For now the record reads as a company that overpaid for growth, was forced to make it pay, and has since chosen its own shares over further empire-building — with the March 2026 debt-funded repurchase as the boldest, and least reversible, expression of that choice.