Summary

  • Adobe spent $2.232 billion on repurchases in Q3 FY2026, equal to 88.5% of its $2.523 billion operating cash flow. The $291 million difference is only an arithmetic remainder, not free cash flow.
  • GAAP net income rose 3.1%, diluted EPS rose 10.5%, and the diluted share count fell 6.8%. Repurchases made the denominator a material part of the per-share result, although Adobe does not attribute the whole share-count move to buybacks.
  • AI-first ARR exceeded $650 million and grew more than 150%, but the disclosed floor is only a little over 2.36% of $27.50 billion total ARR. The next test is absolute conversion into durable contracts and cash, not growth from an undisclosed small base.

The quarter’s largest visible wager was made in cash

Adobe’s Q3 headline is designed to point forward. The company says AI-first ending ARR grew more than 150% from a year earlier, its products crossed one billion monthly active users, and quarterly revenue reached a record $6.760 billion. Those figures describe reach and momentum. They do not identify the quarter’s largest visible allocation of cash.

That figure is $2.232 billion. In the earnings exhibit filed with the SEC, Adobe reports that amount under repurchases of common stock for the three months ended 28 August. Operating activities generated $2.523 billion. Divide one by the other and repurchases absorbed 88.46% of quarterly operating cash.

The subtraction leaves $291 million, but it must not be relabelled free cash flow. Operating cash precedes purchases and sales of investments, capital expenditure, debt repayment and other financing movements. Adobe also repaid $250 million of debt during the quarter. The comparison is useful for one reason: it shows how little of the quarter’s operating inflow remained after the buyback cheque before the rest of the cash-flow statement had its turn.

Adobe says it repurchased approximately 9.5 million shares. Matching that rounded count with the cash-flow line gives about $235 per share. This is a reasonableness check, not the exact price of every execution. Structured arrangements, settlement dates and rounding can separate cash paid from shares delivered. The economic point survives that caveat: capital return was not a marginal use of surplus funds. It was nearly as large as the entire quarter’s cash generation from operations.

The pace was not a one-quarter experiment. Adobe paid $4.590 billion for repurchases in the first half, according to its fiscal-Q2 Form 10-Q. Add the Q3 cash-flow line and nine-month repurchase cash reaches $6.822 billion. That is the mature machine against which the emerging AI revenue stream should be judged.

Per-share growth ran ahead of aggregate profit

The denominator shows up clearly in the income statement. Q3 GAAP net income rose from $1.772 billion to $1.827 billion, an increase of about 3.1%. Diluted EPS rose from $4.18 to $4.62, about 10.5%. The diluted share count used in the calculation fell from 424 million to 395 million, or 6.8%.

This is not proof that buybacks produced every missing share. Diluted counts are weighted averages. Equity awards, employee issuances, option economics and settlement timing also move the denominator. Adobe’s own description of its programme is appropriately broader: return value, minimise dilution from stock issuance and reduce share count over time. What the filed arithmetic does establish is that aggregate profit and per-share profit followed different trajectories, and the lower denominator was material to the gap.

That distinction matters when a company describes double-digit EPS growth alongside an AI transition. Per-share growth can come from a stronger numerator, a smaller denominator, or both. In Adobe’s quarter, revenue grew 13% and operating income 8.3%, while net income grew only 3.1%. Diluted EPS nevertheless grew 10.5%. Investors do not need to choose between “real growth” and “financial engineering” as slogans. They need to keep the two mechanisms separate.

The numerator is the business: pricing, paid seats, usage, retention, costs and margins. The denominator is the number of claims across which that result is divided. Repurchases can be sensible when expected returns on the shares exceed alternative uses of cash. They can also become expensive if the operating thesis disappoints. The filing supplies the cash paid and the share-count movement; it does not supply the counterfactual return on that capital.

AI-first ARR now has a size, and it is still a small layer

Adobe’s prepared remarks offer a better AI signal than the product roll call. AI-first ending ARR exceeded $650 million and grew more than 150% year on year. Total ending ARR was $27.50 billion, up 11.2%.

At the disclosed floor, $650 million is 2.36% of the total. Because Adobe says “exceeded” and both figures are rounded, 2.36% is not the exact mix. It is a scale marker. AI-first ARR is no longer too small to name, but it remains a small measured layer inside the recurring-revenue base that supports the company’s cash generation.

Rapid percentage growth is compatible with that small size. A young line can double or triple without moving the consolidated result as much as a few percentage points of retention or pricing in the mature subscription book. Adobe does not disclose AI-first ARR by product, customer cohort, margin or contract duration. It says Firefly App and credit-pack ARR grew 40% quarter on quarter, but gives no absolute Firefly balance. It says Acrobat AI Assistant monthly active users doubled sequentially, but an active user is not a paid contract.

This is why the one-billion-MAU milestone should be held apart from monetisation. Adobe says monthly active users across its businesses grew more than 20%; the creative freemium group exceeded 100 million and the Business Professionals & Consumers group exceeded 900 million. These are useful funnel measures. They do not reveal conversion rate, price, churn or the incremental cost of serving AI workloads.

The metric hierarchy is therefore important. MAU records reach. ARR annualises the value of recurring contracts at a point in time. RPO records contracted revenue not yet recognised. Revenue is recognised under accounting rules. Operating cash is collected and paid on another timetable. Treating any one of them as a substitute for all the others would make the AI business look more mature than the disclosure allows.

The backlog has a nearer and a later clock

Adobe ended Q3 with $22.16 billion of remaining performance obligations, up 8%. It says 67% was current and that current RPO grew 9%. Multiplying the rounded total by the rounded mix yields about $14.847 billion expected within the current window and $7.313 billion outside it.

Those are analytical allocations, not separately reported balances. They nevertheless show the revenue clock more clearly than the MAU figure. Roughly one-third of the contracted backlog sits beyond the current period. Conversion depends on service delivery, contract terms and customer behaviour; RPO is not cash already in the bank.

The RPO growth rate also trails total ARR growth of 11.2% and subscription revenue growth of 14% in the quarter. That does not by itself signal deterioration. Mix, billing duration, renewals and the timing of large enterprise contracts can move each measure differently. It does mean the next report should be read as a bridge among the metrics, not as a contest over which headline percentage is largest.

For AI-first products, that bridge is still missing. Adobe has disclosed a credible minimum ARR scale and several usage-growth indicators. It has not shown how much of the $22.16 billion RPO belongs to AI-first contracts, what portion is current, or whether AI usage carries a different gross-margin profile from the established applications. Until those fields appear, the young stream cannot be compared with the mature subscription reservoir on equal terms.

The board left the repurchase valve wide open

The capital-return capacity is much less ambiguous. In April, Adobe’s board approved another $25 billion of repurchase authority through 30 April 2030. The authorisation filing says purchases may occur in the open market or through structured arrangements. It also says Adobe has no obligation to spend any amount and will consider market conditions, legal requirements, capital needs and better alternatives.

At the end of Q2, $26.781 billion remained across the March 2024 and April 2026 programmes. By the Q3 call, Adobe said it had fully used the older authority and had approximately $24.55 billion left under the new one. The valve can therefore stay open for years. Authority is not a liability, but it creates a visible choice each quarter between retiring shares and preserving cash for product development, acquisitions, debt reduction or resilience.

The balance sheet records the cumulative pressure without assigning it to a single cause. Cash and short-term investments totalled $5.639 billion at 28 August, down $956 million from the fiscal-year start. Total debt was $6.363 billion, compared with $6.210 billion. Adobe also completed Semrush during the year and moved investments and debt. It would be wrong to charge the entire liquidity change to repurchases. It is still fair to observe that the disclosed buyback cash alone reached $6.822 billion over nine months.

Adobe’s next chief executive inherits this choice. A September Form 8-K says Anil Chakravarthy will become president and CEO on 1 December, with Shantanu Narayen moving to executive chair. The relevant handoff is not a personality contest. It is the relative maturity of two machines: one that can retire billions of dollars of shares now, and one that must turn fast-growing AI usage into a much larger absolute recurring-revenue contribution.

Primary evidence: Adobe’s Q3 earnings exhibit, prepared remarks, investor data sheet, Q2 Form 10-Q, repurchase-authorisation filing and leadership-transition filing.