From Aryan Vatsa | Product & Market Analysis
Adobe's Firefly ARR Rose 40% in a Quarter. Credit Packs Are Working for Someone
On this page
Adobe's Q3 FY2026 release says AI-first ARR grew more than 150% year over year. On the call, management said Firefly ending ARR rose 40% quarter over quarter and AI-first ending ARR passed $650 million. The 8-K defines neither term and gives no dollar base for either growth rate. Credit packs may well be working. The filing cannot prove it.
Key takeaways
- The 150% figure sits on an undisclosed base. Adobe's 8-K states AI-first ARR grew more than 150% year over year, with no dollar amount, no footnote and no list of included products.
- AI-first ARR is about 2.4% of Adobe's $27.5 billion total ARR. The call put it above $650 million, so the fastest-growing line in the release is still a small share of the book.
- Firefly's 40% is a sequential rate on a base nobody printed. Management said ending ARR across the Firefly app and credit packs grew 40% quarter over quarter, and gave no dollar size.
- Credit packs convert usage intensity into recurring revenue, but the evidence is one quarter. The same quarter showed net new ARR down 36% to 37%, according to an analyst on the call.
Reader: founder or director. The question here is what a vendor should disclose about AI revenue, and what to copy from Adobe's credit model, which is a unit economics and pricing decision.
What Adobe actually reported in Q3 FY2026
Start with the primary source. Adobe filed its Q3 FY2026 results as Exhibit 99.1 to an 8-K dated 10 September 2026. The quarter ended 28 August 2026.
Revenue was $6.76 billion, up 13% year over year, or 12% in constant currency. Total Adobe ARR exiting the quarter was $27.50 billion. The full-year target for ending ARR growth stayed at 10.2%.
The headline AI bullet reads: "Adobe AI-first ARR grew more than 150% year over year." That is the only AI revenue sentence in the document. It carries no footnote marker. The only footnotes in the release cover EPS assumptions and the non-GAAP tax rate.
The Firefly figure is not in the 8-K at all. It appears only in prepared remarks on the call. So a reader of the filing alone sees a growth rate for AI-first ARR and nothing about Firefly, credit packs or credit consumption.
| Claim | In the 8-K? | On the call? |
|---|---|---|
| AI-first ARR grew more than 150% year over year | Yes, one bullet, no footnote | Yes, with "exceeding $650 million" |
| AI-first ARR in dollars | No | Yes, as a floor |
| Firefly ending ARR up 40% quarter over quarter | No | Yes, by two speakers |
| Firefly ARR in dollars | No | No |
| A definition of AI-first ARR | No | No |
| Year-ago dollar base for the 150% | No | No |
Compiled from the 8-K exhibit and the Motley Fool transcript, which says it is partly machine-generated and may contain errors. The release points readers to an investor datasheet and slides for more detail. We did not open those, so a definition could sit there.
What "AI-first ARR" counts, and what it does not
ARR means annualised recurring revenue. Companies define it themselves. There is no accounting standard for it, so the definition decides what the number means.
Adobe supplies none for AI-first ARR in either document we read. CEO Shantanu Narayen did offer a contrast on the call. He said investors talk a lot about AI-first revenue, and asked them to think about "AI influence revenue" too, meaning the core Creative Cloud desktop apps. No figure was attached to the second term.
AI-first sounds like products built around AI, such as Firefly and the AI assistants. AI-influenced sounds like existing products that AI makes stickier. The first is measurable as a line. The second is an argument about retention, and you cannot audit it from outside.
That split matters for any vendor reporting an AI number. I would not accept an AI revenue figure from a supplier until I knew which side of that line it sat on.
A secondary summary dated 8 July 2026 reported AI-first ARR above $500 million and Firefly near $300 million for Q2. That page cites aggregators, not an Adobe filing, and does not say what the Firefly figure covers. We use it for context only. No conclusion in this post depends on it.
The missing base: a growth rate is not a size
A growth rate without a base tells you direction, not scale. Take the one dollar figure we do have. If AI-first ending ARR is $650 million, it equals about 2.4% of the $27.5 billion total. That is our arithmetic on two disclosed numbers, not an Adobe statement.
Growth above 150% means at least 2.5 times the prior year. If ending ARR is about $650 million, the year-ago figure was roughly $260 million or less. We cannot be sharper, because $650 million is itself a floor.
That is an implication, not a disclosure. It says the base was small, which is the normal condition of a fast-growing new line. It also says a 150% rate and a 40% quarterly rate describe a line still being built.
The two floors, above $500 million in Q2 and above $650 million in Q3, are not enough to compute a quarterly growth rate. A floor over a floor can hide a gap of any size. Treat the sequence as direction only.
The Firefly 40% claim, read carefully
The wording matters. Anil Chakravarthy, who leads Digital Experience and becomes CEO on 1 December, said Firefly ending ARR "across Firefly app and Firefly credit packs" grew 40% quarter over quarter. Interim CFO Steven Day repeated the same phrase.
Three details limit what you can conclude from it.
The metric covers two things bundled together. It does not say how much comes from monthly Firefly subscriptions and how much from prepaid credit packs. A reader asking whether credit packs specifically are working cannot answer from this sentence.
Forty percent in one quarter is not 40% a year. If it held for four quarters, a base would grow about 3.8 times. Nobody said it would hold, and management gave no forward guide on Firefly.
Forty percent of what? The call does not say. A 40% rise on $100 million and on $400 million are very different stories for Adobe, and the filing leaves both open.
Why credit packs reach ARR when usage is high
The mechanism management described is straightforward. On the call, Chakravarthy said the freemium strategy brings in new users, then Adobe works on "increasing their intensity of AI usage", and that "engagement is what is starting to translate into ARR."
A credit pack is a prepaid block of generation capacity. A user who runs out buys more. That turns heavy use into a purchase event without forcing a seat upgrade.
Chakravarthy said credit consumption was "accelerating quarter on quarter, across Creative Cloud and the Firefly app." Narayen added that video drives a lot of the usage, and that credit consumption inside Creative Cloud was very strong. Neither gave a number.
Consumption and revenue are not the same. Credits that come bundled in a plan are consumed without producing new ARR. Only the purchased kind adds to the line. The call did not separate them.
Adobe says freemium monthly active users passed 100 million, growing more than 70% year over year. The interim CFO defined that base as including Firefly, Express, and web and mobile versions of Premiere, Photoshop and Lightroom. Free users are the pool credit packs sell into.
The cost shows up elsewhere. Remaining performance obligations grew 8%, and the interim CFO tied the slowdown to the freemium model. An analyst on the call put net new ARR down 36% to 37% year over year. Management neither confirmed nor disputed that figure.
What credit packs show other SaaS vendors
I would copy the structure and refuse the disclosure. The structure is sound. The disclosure is the weak part.
Copy: a revenue path that follows usage
Per-seat pricing breaks when one user does the work of five. Credits let revenue track consumption while the seat stays cheap. That is the logic behind the seat-compression pressure covered in our analysis of seat compression and SaaS pricing, and why other vendors have moved the same way. HubSpot's version is mapped in the breakdown of Breeze credits and repricing, and Workday's in the piece on Flex Credits as a seat hedge.
Copy: a free tier that feeds a paid one
Adobe's own explanation is that freemium acquires users and credits monetise the heavy ones. That works only if the free tier has a natural ceiling. A vendor whose free tier never runs out is giving away compute, a trap covered in the comparison of AI design tool credit pricing.
Refuse: a headline rate with no base
If you sell an AI line, publish the dollar figure alongside the growth rate. A rate alone invites exactly the scepticism this post applies to Adobe. Investors who have been burned by loose AI metrics will discount an undefined one, and our ranking of AI revenue disclosure metrics shows which formats survive that test.
| Question | Credit packs | Per-seat pricing |
|---|---|---|
| Revenue follows usage? | Yes, when users buy extra packs | No, a heavy user pays the same |
| Forecastable for the buyer? | Less so | More so |
| Cost of heavy free users | Falls on the vendor until they buy | Low, seats are paid up front |
| Where seats win | Budget approval is easier with a fixed line, and a fixed line is what most finance teams expect. | |
Where this argument is weakest
I am arguing that the disclosure is thin. That is a narrower claim than saying the business is weak, and I want to keep them apart.
The data may exist and we did not open it
The release points to slides and an investor datasheet. A precise AI-first definition and Firefly dollar figures may sit there. I read the 8-K and a full call transcript, not those documents. If they carry the definition, this post overstates the gap.
Our transcript is not an audited record
The call text we used is a Motley Fool transcript, which discloses that it is partly machine-generated and can contain errors. It mis-transcribes at least one analyst's name. The key quotes match across the CEO, CEO-elect and CFO, but treat exact wording as unverified against Adobe's own recording.
The case that this is simply good news
A fair reader can take the other side. A line growing more than 150% to above $650 million inside two years is a real business. Credit consumption accelerating across a 100 million user free base is a sensible place to build a toll. The caution here is about proof, not about whether the model works.
One quarter is not a trend
Net new ARR fell as freemium expanded, and Narayen said he was glad Adobe did not pursue price rises that "may have provided some short term relief". Whether credit packs eventually fill that gap is a question for the next four quarters, not this one.
What it means if you buy or sell AI credits
You are probably not Adobe. The lessons shrink to contract terms and to your own reporting.
As a buyer, ask what a credit buys in a unit you can verify, such as an image, a minute of video or a document. Ask whether bundled and purchased credits expire differently. Ask what happens to unused packs at renewal. A vendor who cannot answer in one page has not priced the product yet.
As a seller, report the base next to every rate. If a metric is new, define it in a footnote the first time it appears. My view is that a new AI revenue metric with no definition costs more trust than the growth rate earns.
Frequently asked questions
What is Adobe's AI-first ARR?
Adobe's Q3 FY2026 release says AI-first ARR grew more than 150% year over year, and the CEO said on the call it exceeded $650 million. The release gives no definition, footnote or list of included products. Treat it as Adobe's own metric, not one tied to an accounting standard, until the company publishes its construction.
How much did Adobe Firefly ARR grow in Q3 2026?
Management said Firefly ending ARR across the Firefly app and Firefly credit packs grew 40% quarter over quarter. That is a sequential rate, not an annual one. No dollar size or prior-quarter base was given on the call, and the figure does not appear in the 8-K itself.
Are Adobe's AI credit packs working?
The evidence is encouraging but thin. Firefly ARR rose 40% in a quarter and credit consumption accelerated, according to management. But the metric combines the app and the packs, no dollars were disclosed, and net new ARR fell in the same quarter. One quarter cannot separate pack revenue from subscription revenue.
Why did Adobe stock fall despite record revenue?
We could not verify a single cause. Adobe posted record Q3 revenue of $6.76 billion, yet an analyst cited net new ARR down 36% to 37% and RPO growth slowed to 8%. A secondary report put the shares down about 27% for the year, but we could not open that source.
What does Adobe's credit pack model show other SaaS vendors?
It shows that prepaid credits can tie recurring revenue to usage intensity while a free tier feeds the funnel. It does not yet show that this offsets slower seat-based growth. Vendors should copy the structure, publish the dollar base beside every growth rate, and define any new AI metric before reporting it.
How should I read an AI ARR growth figure?
Ask for three things: the definition, the dollar base and the period. A rate like 150% or 40% means little without them. Check whether it is year over year or sequential, and whether it includes bundled usage or only purchases. If a vendor cannot supply all three, treat the number as marketing.
Where to start this week
Open your own pricing page and find the one AI number you publish or quote. Write its definition, its dollar base and its period beside it. If you cannot, that gap is the first fix.
Then read Adobe's investor datasheet for the AI-first definition and the Firefly dollar figure. If they are there, update your view of this post. If they are not, you have your answer about how much to trust the rate.
Related on this topic
For how Microsoft meters AI usage, see the map of Copilot credits and metered billing. For the wider choice between flat and metered models, see bundling versus metering for AI features.
References
- Adobe Inc., Form 8-K, Exhibit 99.1, Q3 FY2026 results, 10 September 2026. Used for revenue, total ARR, targets, and the AI-first ARR sentence.
- Motley Fool, Adobe Q3 2026 earnings call transcript, 11 September 2026. Used for the $650 million floor, the Firefly 40% statement, freemium MAU, RPO and the analyst questions. Partly machine-generated.
- SaasRise, Adobe AI revenue tops $500M as Firefly nears $300M ARR, 8 July 2026. Secondary summary; Q2 context only.
Weakest part of the source base: the Firefly and AI-first dollar figures rest on one machine-assisted transcript and one aggregator. The investor datasheet and slides were not opened. Figures current as of 14 October 2026. The chart showing the credit pack path is illustrative.
Related reading