Adobe is not being killed by artificial intelligence. It is being made more expensive to run by it — and those are very different problems with very different price tags.
We published a full X-Ray of Adobe on 31 July 2026 at $250.41, arguing the shares were priced for a melt the numbers did not show. Seven weeks later, on 16 September, we revisited it after the third-quarter results. The price was $250.50 — nine cents higher. In between, the company named a new chief executive, lost the man who ran its creative business, and published a quarter that quietly retired the old argument.
"Nobody will need Photoshop"
It was a good story. The accounts never supported it.
The case against Adobe was elegant: generative models make images from a sentence, so the expensive tool that makes images from skill becomes obsolete, and a franchise built over thirty years evaporates in three.
The market believed it thoroughly. From the 2021 peak to our July report the shares fell about 64%. Over the same stretch, revenue rose about 60% and earnings per share about 74%. Gross margin reached an all-time high of 89%. Return on invested capital was 36%. A company being destroyed does not usually post its best-ever margin on the way down.
What actually fell was the multiple. The business kept compounding; the market stopped paying for it. By July the shares changed hands at roughly fourteen times earnings and ten times free cash flow — cheaper than Salesforce, with fifteen points more operating margin.
A 64% fall while profits rise 74% is not a verdict on the business. It is a verdict on its future, cast in advance.
Revenue accelerated. The order book did not.
Q3 FY2026, reported 10 September
| Measure | Q1 FY26 | Q2 FY26 | Q3 FY26 |
|---|---|---|---|
| Revenue growth | +12.0% | — | +12.9% |
| Total recurring revenue growth | +12.5% | — | +11.2% |
| Remaining performance obligations | +13% | +13% | +8% |
| Net new recurring revenue | $0.40bn | $1.04bn (incl. ~$0.48bn bought with Semrush) | $0.40bn |
Read those two lines together, because separately they each tell a lie. Revenue is what Adobe billed; remaining performance obligations are what customers have contracted and not yet been billed for. One looks backwards, one looks forwards. Revenue accelerating while the forward book decelerates from 13% to 8% — and shrinks in absolute terms — is the shape of a company collecting well on commitments made earlier, while making fewer new ones.
The net new recurring revenue line says the same thing without the jargon. Q3 added the same $0.40 billion as Q1 — two quarters later, with fifty million more free users in the funnel.
The bill, not the moat
A margin that had not moved in eight quarters, moved
Here is the number that matters more than any of them. Adobe's gross margin came in at 88.1%, down 1.2 points — its first break out of an eight-quarter band. For a software company, a gross margin is not supposed to have a story. This one does.
Firefly, Adobe's AI layer, now serves other companies' models inside Adobe's own products — Gemini, GPT-Image, FLUX, Runway, Kling, Topaz. Every generation runs on somebody's hardware, and Adobe pays for that inference. Often it gives the result away, because the strategy is to put a free creative tool in front of as many people as possible: monthly active users reached one billion, and free creative users doubled from 50 million to 100 million in a year.
And the part that pays? AI-first recurring revenue of $650 million, growing 150% — 2.4% of a $27.5 billion book, after roughly three years. Adobe also stopped updating the far larger "AI-influenced" figure it once led with. When a company retires its favourite number and starts reporting a stricter one, it is usually telling you which of the two it believes.
Give the tool away to a hundred million people, pay the inference cost yourself, and convert enough of them later to justify it. It is a defensible strategy — it is roughly what Adobe did to the boxed-software era. But the payoff was explicitly deferred to 2027, by an executive who will not be there to deliver it: David Wadhwani, who ran the creative business, left. The new chief executive, Shantanu Chakravarthy, comes from the enterprise side and has never run Creative Cloud.
Everyone is paying the same bill
Which is how you know it is the industry, not the company
| Growth | What it cost them | |
|---|---|---|
| Adobe | Revenue +12.9% | Gross margin 88.1%, down 1.2 pts · operating margin 34.8% · profitable |
| Figma | Revenue +48% | Cost of revenue +117% · GAAP operating margin −32% · shares −57% in a year |
| Canva | Outlook cut by a third, to ~20% | Cited the cost of frontier models · marked down 17% by its backers, $42bn to $34.9bn |
This table is the argument. If AI were destroying Adobe and handing the market to nimbler rivals, the rivals would be getting richer. Instead the fastest grower in the category is burning a third of its revenue at the operating line, and the private one just got marked down by the people who own it.
Generative AI did not redistribute the profits in creative software. It raised the cost of being in it. Adobe is the only one of the three paying that bill out of a 34.8% operating margin — which is a real advantage, and a much smaller one than the 89% gross margin used to imply.
The question stopped being whether these tools survive. It is what they earn.
A duller risk, and a cheaper price
We cut Adobe's overall score from 7.4 in July to 6.8 in September, and the cut was concentrated in two dials: the moat, and management. Not because the franchise cracked — a billion monthly users and $27.5 billion of recurring revenue is not a cracked franchise — but because the economics of defending it got worse, and the people who knew how to defend it left.
What replaces the old bear case is something less dramatic and more likely: Adobe stays essential, keeps growing at low double digits, and earns a bit less on each dollar than it used to. That is an ordinary outcome for an extraordinary business. At roughly nine times free cash flow, with about 7% of the shares retired in a year and no dividend, an ordinary outcome is already more than the price requires.
The risk we would actually watch is not Canva. It is the forward book. If remaining performance obligations spend another two quarters at 8% while free users keep climbing, then the freemium bet is not converting, and the 2027 payoff becomes a story rather than a plan.
Both Adobe X-Rays — July's call, and September's scorecard of it.
One that turned the AI bill into revenue, and one whose moat was genuinely tested.
What to take from this if you own it — or don't
The general lesson outlives the ticker. When a story about technological extinction meets a set of accounts, check which line the story actually shows up in. For three years the Adobe bears pointed at the share price, which is not evidence. The first real evidence arrived in September, and it arrived in the gross margin — a line nobody was watching, in a quarter where the headline numbers were fine.
That is usually how it goes. The story is loud and the evidence is boring, and the evidence is in a different line than the one everyone is arguing about.
Frequently asked
Is AI destroying Adobe's business?
Not on the evidence to September 2026. Revenue grew 12.9% in the third quarter of fiscal 2026 and total recurring revenue reached $27.5 billion. What changed is cost, not demand: gross margin fell to 88.1%, its first break in eight quarters, because Adobe now pays for inference on other companies' models inside its own product.
Why has Adobe stock fallen so far if the numbers are fine?
Because the market is pricing a future, not a quarter. Since 2021 Adobe's revenue is up about 60% and earnings per share up about 74%, while the shares fell roughly 64% from the peak. Almost the entire decline is multiple compression — the market paying less for each dollar of the same profit, on the belief that those dollars will not last.
How much money is Adobe actually making from AI?
Its AI-first recurring revenue was $650 million as of the third quarter of fiscal 2026, growing 150% — and equal to 2.4% of a $27.5 billion book after roughly three years of effort. Adobe also stopped updating the far larger 'AI-influenced' figure it used to lead with, which tells you which of the two it now considers meaningful.
Are Canva and Figma winning instead?
They are growing faster and paying for it. In 2026 Figma grew revenue 48% while its cost of revenue rose 117%, giving a GAAP operating margin of about −32%, and its shares fell 57% in a year. Canva cut its growth outlook by roughly a third and was marked down 17% by its own backers, citing the cost of frontier models. The AI bill is arriving for everyone; Adobe is the one still paying it out of a 34.8% operating margin.