Capital Allocation

Capital Allocation

What management does with Cognizant's cash is the bridge between a cheap cash generator and the value that reaches a shareholder. Over 2021–2025 the company returned about $8.1 billion — roughly three-quarters of its free cash flow — split between buybacks and a growing dividend, all on a net-cash balance sheet. The development that matters now is how it is deploying that policy into the derating. Cognizant doubled its 2026 buyback to $2 billion with the stock in the low $40s, ~20% below a ~$55 no-growth free-cash-flow floor, so at ~$44 the program retires about 9.5% of shares a year versus 5.5% at the $75.40 it paid a year earlier — but directors and officers as a group hold only ~0.43% of the stock (~0.13% vested), with a single ~$210k open-market purchase across ~398 Form 4 transactions.

Both halves belong in the same breath. The accretion is real only if $44 is genuinely below intrinsic value: Margin of Safety puts the no-growth floor near $55, so the same $2 billion the board added to the 2026 program [1] retires roughly 9.5% of the shares — compounding per-share value for a non-seller, but amplifying the durability bet rather than hedging it. In the bear case, where the true no-growth value sits below $44, the same program destroys per-share value by overpaying with the company's own cash. The insider record is the counter-fact a fallen-star screen tests directly: a management group that owns only ~0.43% of the stock [2], with a single ~$210,000 open-market purchase across roughly 398 Form 4 transactions (Ownership and Governance develops the ownership record), is not the conviction signal that usually accompanies a mispriced-cheap thesis. On the reader's fallen-star-with-strong-insider-ownership screen, this leg fails.

Nearly all the cash goes back out

Cognizant runs to a stated capital-allocation framework: deploy roughly 100% of free cash flow each year, targeting about 50% for acquisitions, 25% for share repurchases and 25% for dividends [3]. That split has been the policy for years — the same 50/25/25 language appears in the 2021 10-K [4], Capital Allocation Framework — p.60").

In practice, the shareholder-return half of that policy is steady and large. Over 2021–2025 Cognizant paid out $5.24 billion in buybacks and $2.87 billion in dividends — $8.1 billion returned, against $10.9 billion of free cash flow generated over the same five years, or about three-quarters of it. The rest went to acquisitions, and the balance sheet still ended in a net-cash position. This is a company that neither hoards cash nor stretches for it.

Source: derived from FY2021–FY2025 10-Ks, Consolidated Statements of Cash Flows [5], Consolidated Statements of Cash Flows — p.90"). Free cash flow = operating cash flow less capital expenditure.

Source: derived from FY2021–FY2025 10-Ks, cash-flow statements, balance sheet and share counts [6], Consolidated Statements of Cash Flows — p.90").

The dividend is the quieter half. Cognizant only initiated a dividend in 2017 — late for a company of its age — and has raised it every year since, paying $1.24 per share in 2025 and lifting the quarterly rate to $0.33 in February 2026 [7]. At around $44 that is close to a 3% yield, but it is the smaller lever; the buyback and M&A move the needle.

The framework flexes — 2024 was the acquisition year

The 50/25/25 split is a target, not a rule, and the "50% for acquisitions" slice is the one that swings hardest. Most years buybacks dominate; in 2024 the mix inverted. Cognizant cut repurchases to $605 million — less than half its usual pace — and spent the room on deals: Thirdera, a ServiceNow partner, for $428 million in January, and Belcan, an aerospace and defense engineering-services firm, for $1.36 billion in August ($1.20 billion cash plus 1.47 million shares) [8]. On the company's own accounting, 2024 ran 57% to acquisitions and only 22% to buybacks [9]. The pattern continued into 2026 with 3Cloud, a Microsoft Azure and AI-services firm, for $733 million [10].

The acquisitions are where a skeptic's attention is best spent. A decade of deals has lifted goodwill from about $2.6 billion in 2016 to $7.1 billion at the end of 2025 — now roughly 47% of shareholders' equity [11]. None of it has been impaired, which is a point in management's favor, but the deals' contribution is hard to separate from the core. Belcan is the clearest case: it anchored the 2024 acquisitions that added $384 million of revenue in their first partial year and lifted 2025's reported growth to 7% [12] — the same reported figure that, as Standing Among Peers showed, still left Cognizant with the second-lowest multi-year growth in its peer group once the acquisitions are stripped out. The acquisitions buy growth and adjacency; whether they compound value per share above their cost is, so far, unproven.

Source: derived from FY2021–FY2025 10-Ks, Consolidated Statements of Cash Flows [13], Consolidated Statements of Cash Flows — p.90"). The 2024 dip funded the Thirdera and Belcan acquisitions.

The buyback has been mechanical, not shrewd

The buyback is real money — but its timing has been price-insensitive rather than opportunistic, and that matters to a value investor for whom the case is most sensitive to buying cheap. The clearest evidence is the fourth quarter of 2025, when Cognizant repurchased $325 million of stock at an average of $75.40 a share, buying more as the price climbed — 0.7 million shares in October at $67.58, then 2.1 million in November at $72.81, then 1.5 million in December at $82.90 [14], Cash Dividends and Issuer Purchases — p.48"). Within roughly six months the stock had fallen into the low $40s. Management was not buying its stock because it was cheap; it was buying on a schedule, and the schedule ran straight into a top.

That price-insensitivity also explains why five years and $5.24 billion of repurchases shrank the diluted share count only about 7% — from 528 million in 2021 to 489 million in 2025, under 2% a year. Buying at $70–90 for most of the period, with stock-based pay and deal shares leaning the other way, blunted the per-dollar effect. Steady, yes; shrewd, no.

The derating changes the arithmetic

Here the story turns. In May 2026 — with the stock already halved — the board added $2 billion to the repurchase authorization and, more tellingly, raised the amount it expects to repurchase in 2026 from $1 billion to $2 billion, leaving about $3.45 billion authorized [15]. First-quarter 2026 repurchases had already stepped up to $444 million, more than double the $209 million of a year earlier. The same buyback that bought high in 2025 is now being pointed, at scale, at a much lower price.

The arithmetic of a fixed-dollar buyback is unforgiving in the company's favor here. The identical $2 billion retires far more stock at $44 than at the $75.40 it paid a year earlier:

Source: derived from the 2026 repurchase target [16] and Q4 2025 average repurchase price [17], Cash Dividends and Issuer Purchases — p.48"); shares out ~479M.

At $44, a $2 billion program retires roughly 9.5% of the shares outstanding in a single year — nearly double the 5.5% the same spend would have bought at last year's prices. For a holder who is not selling, the derating and a maintained buyback compound in their favor: the cheaper the stock stays, the more each dollar of the company's own cash buys back. The honest caveat is that this only helps if the shares are genuinely cheap rather than cheap for a reason — the durability questions the earlier chapters raised — and that management's demonstrated willingness to buy at any price cuts both ways.

The balance sheet removes the tail risk

For a reader who will not own anything that could go bankrupt, the financing side closes cleanly. At the end of 2025 Cognizant held $1.9 billion of cash and short-term investments against just $576 million of total debt — $33 million short-term and $543 million long-term — a net-cash position of roughly $1.3 billion [18][19]. The capital-return program is funded entirely by operating cash, not leverage, and an untapped revolving facility sits behind it. Whatever the outcome of the AI-durability debate, solvency is not part of it.

What would change the read

The capital-allocation case is a qualified positive: the cash comes back, the balance sheet is unlevered, and the derating has quietly turned a mediocre buyback into a materially more accretive one. Three things would sharpen — or spoil — that read. First, whether management actually executes the raised $2 billion at these prices, rather than tapering when the stock recovers; the 2024 pivot shows the framework bends toward deals when management prefers them. Second, whether the 3Cloud and Belcan acquisitions earn their $7.1 billion of goodwill, or whether a future impairment reveals the M&A slice as growth bought too dearly. Third, whether the buyback's per-share bite finally widens now that the price is low — a share count that keeps shrinking only ~2% a year, even at $44, would signal the cash is leaking to dilution and deals faster than it returns to owners.