Adobe AI-First ARR Grew 150 Percent. The Stock Fell Anyway.

Adobe has spent two years being treated as generative AI's most obvious casualty, a design-software near-monopoly about to be commoditized by free image models. On Wednesday evening it delivered its cleanest numerical rebuttal yet, and the market sold the stock anyway.

Fiscal third-quarter revenue hit a record $6.76 billion, up 13 percent as reported and 12 percent in constant currency, ahead of a consensus near $6.69 billion. Non-GAAP earnings per share came in at $6.13, up 15 percent and six cents above estimates. And the metric the bulls had circled finally moved: AI-first ending annual recurring revenue now exceeds $650 million, growing more than 150 percent year over year. Shares closed the regular session at $248.95, down 2.32 percent, then shed another 2.14 percent after hours to $243.50 — a cumulative 4.45 percent decline on a quarter Adobe beat and a year it raised.

The sell-off has two sources, and neither is the AI number. The first is guidance arithmetic. Adobe guided fourth-quarter revenue to $6.80 billion to $6.85 billion, a midpoint of about $6.825 billion, short of consensus, with non-GAAP EPS of $6.30 to $6.35 and an operating margin near 44 percent. Full-year targets went up to $26.576 billion to $26.626 billion in revenue and $24.45 to $24.50 in non-GAAP EPS — a raise of roughly $50 million at the midpoint, slightly less than the quarter's own beat. Interim CFO Steve Day attributed the shortfall entirely to currency. “Nothing changing in the business at all,” he told Citi's Andrew Gerard. “There is a slight FX headwind that we are having coming into Q4, and it is just netting that off.”

The second source is harder to wave away. Total ending ARR reached $27.5 billion, up 11.2 percent — decelerating. Remaining performance obligations came in at $22.16 billion, up 8 percent, which Griffin Securities analyst Jay Vleeschhouwer flagged on the call as the first single-digit RPO growth since early fiscal 2023. Evercore ISI put net new ARR down 36 to 37 percent year over year. Management says this is deliberate: Adobe is pushing traffic into a freemium funnel rather than converting it, and deferred planned Creative Cloud price increases to do so. The engagement numbers support the story. Monthly active users across all businesses crossed 1 billion, growing more than 20 percent. Acrobat plus Express MAU passed 900 million, up more than 25 percent. Creative freemium MAU crossed 100 million and grew more than 70 percent. Acrobat AI Assistant MAU doubled quarter over quarter, and Firefly ending ARR across the app and credit packs rose 40 percent sequentially.

Outgoing chief executive Shantanu Narayen was unapologetic about the trade. “I'm actually really happy that we didn't focus on the pricing actions,” he said, “because that, while it may have provided some short-term relief, would not be as critical as continuing to drive new user adoption.”

All of it lands in the middle of a leadership handoff. Anil Chakravarthy, president of customer experience and orchestration and formerly chief executive of Informatica, becomes Adobe's president and CEO on December 1; Narayen, who announced his exit in March after 18 years in the job, becomes executive chair. “I am deeply honored to lead Adobe at this pivotal moment,” Chakravarthy said, framing his mandate around what he called the era of agentic software.

A 150 percent growth rate on 2.4 percent of the business

Here is the number that explains the share price. AI-first ARR of $650 million against total ending ARR of $27.5 billion is 2.4 percent of Adobe's book of business. Mizuho made exactly this point when it downgraded the stock in April, noting the AI-first line was then under 2 percent of the base. Triple it again and it is still only around 7 percent. Meanwhile creative and marketing professional subscription revenue — the business that is supposedly under existential threat — was $4.65 billion in the quarter alone, growing 13 percent.

So 150 percent is real evidence and thin evidence at once. It proves Adobe can charge for AI. It does not prove Adobe is winning against Google's Nano Banana family, which Sundar Pichai said had passed 50 billion cumulative images by May, or OpenAI's image models, both effectively free at consumer scale.

The disclosure choice is itself the tell. AI-first ARR is a basket: the Firefly app, Firefly credit packs, Firefly Enterprise, Acrobat AI Assistant, GenStudio, and the new brand-visibility products built on Semrush. Those monetize through credits, seats and enterprise contracts, serve different buyers, and face different competitors. Aggregating them produces a flattering growth rate and conceals the one line investors actually want — standalone Firefly versus the free alternatives. Narayen came close to conceding the framing was incomplete. “I know we talk a lot about the AI-first revenue,” he said. “If you think about the AI-influenced revenue and what is being driven by Creative Cloud, all of the core desktop applications, the credit consumption there also is very robust.” That is the more honest bull case, and Adobe does not report it.

Adobe is betting on distribution rather than model supremacy, extending Acrobat into ChatGPT, Chrome, Microsoft Edge and WhatsApp and routing tasks to whichever third-party model performs best. It is a deliberate choice to be the application layer on top of commodity intelligence — defensible, but it caps how much AI margin Adobe ever captures.

What to watch

Four things. Adobe MAX in Miami Beach in November, Chakravarthy's first, where the Firefly and Creative Cloud roadmap either justifies the freemium patience or does not. Whether the Topaz Labs acquisition closes in the fourth quarter as guided. Whether the seasonally strong Q4 delivers the step-up in net new ARR required to hit the reiterated 10.2 percent full-year book-of-business growth. And above all whether Chakravarthy reinstates the deferred Creative Cloud pricing — the single clearest signal of whether Adobe believes those 100 million freemium creators will pay.