Adobe's AI Growth Still Runs on the Installed Base
Executive summary
Adobe reported a strong fiscal third quarter on September 10: revenue rose 13% to $6.76 billion, customer-group subscription revenue rose 14% to $6.56 billion, and total annualized recurring revenue reached $27.50 billion. The company also passed one billion monthly active users across its creativity and productivity products. Its AI-first ARR exceeded $650 million and grew more than 150% year over year. Source: Adobe Q3 FY2026 earnings release
The mainstream interpretation is straightforward: Adobe has found its AI growth engine. The evidence supports part of that claim. Direct AI revenue is growing quickly, Firefly usage is deepening, and subscription growth has accelerated without sacrificing the company's high operating profitability.
What that interpretation misses is scale. AI-first ARR is still less than 2.4% of total ARR when the disclosed lower bound of $650 million is compared with $27.50 billion. Even a 150% growth rate on that base cannot yet explain most of Adobe's subscription performance. The core products, installed customer relationships, packaging changes and enterprise distribution still carry the business.
Our contrarian thesis is not that Adobe is losing the AI transition. It is that the transition should be judged as a conversion system rather than a feature race. Adobe is using free products, embedded AI, external distribution and enterprise agents to widen usage. The decisive question is whether those routes create incremental paid relationships and net retention faster than generative tools erode legacy value pools such as standalone stock content.
Market context
Adobe entered this transition with unusual protection. In fiscal 2025, subscription revenue was $22.90 billion, or 96% of total revenue, while total ARR ended at $25.20 billion. Creative workflows, document standards and enterprise marketing systems create recurring use and meaningful switching costs. The installed base gives Adobe time to redesign products before new AI-native competitors become complete workflow replacements. Source: Adobe fiscal 2025 Form 10-K
That protection also makes the evidence harder to interpret. When AI is added to Creative Cloud or Acrobat, revenue from an existing subscription can be described as AI-influenced even if the customer did not pay more or would have renewed without the feature. Adobe said AI-influenced ARR exceeded one-third of its book exiting fiscal 2025. That is evidence of product reach, but it is not the same as direct monetization.
Adobe's narrower AI-first measure covers products built around AI as the primary experience, including offerings such as Firefly and Acrobat AI Assistant. This is a more demanding commercial signal. It passed $500 million in Q2 and $650 million in Q3. Management also disclosed that creative freemium users exceeded 100 million, up more than 70% year over year, while total monthly active users exceeded one billion.
Those figures describe the top and bottom of a funnel without revealing the middle. Public disclosures do not show free-to-paid conversion by product, gross retention for AI-first cohorts, revenue per active user, generative-compute cost per paid account or the portion of Creative Cloud price realization attributable to AI.
Findings
Finding 1
AI-first ARR is growing much faster than Adobe, but it is not yet large enough to carry Adobe.
| Fiscal period | Revenue | Customer-group subscription revenue | Total ending ARR | AI-first ARR | Monthly active users |
|---|---|---|---|---|---|
| Q1 FY2026 | $6.40bn | $6.17bn | $26.06bn | More than tripled year over year; dollar value not disclosed | More than 850m |
| Q2 FY2026 | $6.62bn | $6.39bn | $27.10bn, including about $480m from Semrush | More than $500m; about 3x year over year | Not disclosed in the release |
| Q3 FY2026 | $6.76bn | $6.56bn | $27.50bn | More than $650m; more than 150% year over year | More than 1bn |
Sources: Adobe Q1, Q2 and Q3 FY2026 earnings releases and prepared remarks, published March 12, June 11 and September 10, 2026. Units are U.S. dollars and monthly active users. ARR is a point-in-time annualized measure; revenue is for the quarter. Q2 total ARR includes approximately $480 million acquired with Semrush, so sequential movement is not organic. Values described as "more than" are company thresholds, not exact amounts.
At the Q3 disclosed floors, AI-first ARR represented 2.36% of total ARR. Because Adobe said both figures exceeded their thresholds, the exact share is unknown. The calculation is useful only as a scale marker: direct AI products remain a small part of the recurring base.
This matters because a fast-growing small product and a healthy incumbent can coexist without the first causing the second. Total ARR grew 11.2% year over year in Q3. Customer subscription revenue grew 14%. These are strong results, but the public numbers do not permit the conclusion that AI-first products produced most of that growth.
The installed base is doing strategic work. Adobe can place AI inside products customers already use, bundle capabilities into higher-value plans, sell credit packs to heavy users and introduce enterprise automation through existing procurement relationships. That distribution advantage is real. It also means investors and operators need to separate defensive AI, which helps retain or reprice the core, from incremental AI, which wins new revenue that would not otherwise exist.
Finding 2
The billion-user milestone measures reach, not subscription quality.
Adobe's total monthly active users grew more than 20% year over year in Q3. Creative freemium users crossed 100 million and grew more than 70%; the Business Professionals and Consumers group exceeded 900 million users, up more than 25%. Acrobat AI Assistant users doubled quarter over quarter. These figures show that Adobe can distribute new experiences at enormous scale. Source: Adobe Q3 FY2026 prepared remarks
The same metrics can improve while unit economics weaken. Free users consume storage, model inference, support and product-development capacity. They may attract advertisers or improve brand reach, but Adobe's model remains overwhelmingly subscription-funded. A large free audience becomes valuable when it produces paid conversion, lowers acquisition cost, creates enterprise pull-through or improves retention among existing subscribers.
Adobe acknowledged the timing issue in Q1, saying rapid freemium growth dampened ARR in the short term while building a foundation for later monetization. That is a rational investment, not evidence of failure. It also creates a testable obligation. Conversion should appear in cohort data over time rather than remain inferred from aggregate user growth.
| Funnel stage | Public evidence | Missing economic test |
|---|---|---|
| Reach | More than 1bn total MAU in Q3 | Unique users by product and overlap across surfaces |
| Free creative use | More than 100m MAU, up over 70% | Conversion by acquisition channel and cohort |
| AI engagement | Credit consumption accelerated quarter over quarter | Compute cost per active and per paid user |
| Direct monetization | AI-first ARR above $650m | Gross retention, expansion and contribution margin |
| Core protection | Total ARR up 11.2%; subscription revenue up 14% | Renewal and price realization with versus without AI use |
Source: Blackrock Research framework using Adobe Q3 FY2026 disclosures. Period: quarter ended August 28, 2026. Units are users, growth rates and U.S. dollars as labeled. The final column identifies undisclosed measurements and should not be read as observed company data.
Adobe's distribution through ChatGPT, Chrome, Claude, Edge and WhatsApp complicates the funnel further. External surfaces can expand discovery and make Adobe capabilities easier to invoke. They can also weaken direct customer ownership if the conversational platform controls identity, defaults and the starting interface. Adobe needs to know whether an externally initiated task becomes a durable Adobe relationship or remains an occasional utility call.
Finding 3
AI monetization is partly a replacement problem inside Adobe's own portfolio.
The transition is not purely additive. In Q1, management said the traditional standalone Stock business declined faster than expected as customers gained a choice between licensed stock assets and generated content. That is a clean example of AI shifting value inside an incumbent rather than creating entirely new demand.
Firefly can recapture part of that displaced spend through subscriptions, credit packs and enterprise services. Q3 Firefly ending ARR across the app and credit packs grew 40% from Q2, while Adobe reported accelerating enterprise usage. Yet the gross economics differ. A stock license monetizes an existing asset; a generated image or video invokes models and infrastructure. Revenue substitution is not automatically margin-neutral.
The same issue applies to productivity. An Acrobat AI feature may support a plan upgrade, reduce churn or create a new subscription. It may also become an expected capability whose cost is absorbed into the existing price. Operators often label all four outcomes "AI adoption," though they have different effects on lifetime value.
This is why the one-third AI-influenced share can be directionally useful but economically ambiguous. Embedded AI may be essential to defend the franchise even when it produces no separate line of revenue. Conversely, an AI-first subscription can be incremental but small. The right portfolio view measures both: revenue protected in the core and revenue created outside it, net of cannibalization and delivery cost.
Implications for operators
Subscription companies should define AI revenue with a hierarchy. Separate direct AI subscriptions and usage charges from plan upgrades, retention influence and simple feature exposure. Do not add these categories together as if they represent the same dollar.
Manage freemium as a set of cohorts, not a user milestone. Track the source that acquired each user, the first valuable workflow completed, 30- and 90-day retained use, conversion timing, paid revenue, compute cost and support burden. A billion users can conceal many low-intent visits; a smaller cohort completing repeated professional work may be worth more.
Measure cannibalization explicitly. If a new generative product replaces an older asset, seat or service, compare combined customer contribution before and after migration. Product teams should receive credit for preserving a relationship, but finance should distinguish preservation from net-new growth.
Price around economic value and cost-to-serve. Credit packs can align high model usage with revenue, while bundled allowances reduce trial friction. The balance should vary by modality because video, audio, images and text impose different compute costs and produce different customer value.
Finally, treat external AI platforms as distribution channels with channel economics. Track conversion, repeat use, identity capture, revenue share where relevant and dependence on platform defaults. Integration breadth is helpful only if it expands Adobe's customer relationship rather than making the product interchangeable behind somebody else's interface.
Risks & open questions
This thesis would be falsified if Adobe discloses that AI-first cohorts are already generating a disproportionate share of net new ARR with attractive gross margins, or that AI use materially lifts renewal and expansion after controlling for customer size and plan. It would also weaken if free users convert at rising rates without a corresponding increase in acquisition or inference cost.
The reverse risk is that AI-first growth decelerates before reaching material scale, while generative tools continue to pressure Stock or reduce the differentiation of flagship applications. Adobe's strong installed base can delay the financial evidence of disruption; it cannot eliminate the need to earn preference in new workflows.
Semrush also clouds sequential comparisons. Q2 total ARR included approximately $480 million from the acquisition. Adobe's Q3 total ARR growth is reported year over year and remains useful, but quarter-to-quarter changes should not be treated as a clean organic series.
Management transition adds execution risk. Anil Chakravarthy is scheduled to become chief executive on December 1. The next leadership team must balance rapid free distribution, core subscription packaging, model partnerships, enterprise sales and the cost of running generative services without confusing activity for monetization.
Appendix / methodology notes
This report uses Adobe's fiscal 2025 Form 10-K; fiscal Q1, Q2 and Q3 2026 earnings releases; and Q1 and Q3 prepared remarks. Fiscal Q3 ended August 28, 2026. Company-reported non-GAAP measures are not used to estimate missing values.
The 2.36% scale marker divides the disclosed minimum AI-first ARR of $650 million by total ending ARR of $27.50 billion. Because AI-first ARR was described as exceeding $650 million, the actual ratio is higher; because the exact figure was not disclosed, the calculation must not be presented as an exact share.
ARR is not quarterly revenue and MAU is not a subscriber count. The table places them together only to show scale and progression, not to imply direct conversion. Q2 includes acquired Semrush ARR and is labeled accordingly.
The most useful next chart would follow acquisition cohorts from free activation through paid conversion and renewal. Required inputs are monthly new users by source, retained users at 30/90/365 days, first paid date, direct subscription and credit revenue, inference and storage cost, gross churn, expansion and displaced legacy revenue. Adobe does not publish these inputs. Without them, the evidence supports rapid reach and early direct monetization, but not yet the claim that AI-first products carry the economics of the whole subscription base.