Adobe Stock Analysis: Record Revenue, a 69% Crash and the Pivot Beyond Creative Software

Adobe is one of the most successful software companies ever built, and for the last two years the market has been pricing it like a company with a terminal illness. Both of those statements are supported by the same set of numbers, which is what makes this interesting.

The usual story told about Adobe is that generative AI is eating it alive. That story is wrong in an interesting way. AI has not taken Adobe's revenue. What it has taken is Adobe's pricing power and its entry point, and the company's response has been to stop defending the thing it was famous for and start buying the thing nobody owned yet.

This is a long look at what actually happened, built on Adobe's own filings and earnings calls rather than on the narrative. Where the primary sources contradict the commentary, the primary sources win, and I have flagged every place they disagree.

The Fall: Two Crashes, Not One

People talk about the Adobe crash as a single event. It was two, with a full recovery in between, and the shape matters because the two falls had completely different causes.

The first crash, from $699.54 in November 2021 to $275.20 in September 2022, was a 60.7% fall and it was mostly not about AI at all. It was the rate-hike repricing of every long-duration growth asset on the market, plus a self-inflicted wound: Adobe announced its $20 billion bid for Figma on 15 September 2022, and the market hated it immediately.

Then the stock recovered 124.5% to $617.78 by January 2024, back to within 11.7% of its all-time high. And then it fell 69.2% to $190.12 by June 2026.

The reason for the second decline was not one problem. It was a war on five fronts at once.

The Escalation: Image Models Did Not Break Adobe. Video Did.

The single most useful thing in this whole story is a date, and almost nobody puts it in the right place.

ChatGPT launched on 30 November 2022. Adobe's stock went up. The image generation wave, DALL-E 2, Midjourney and Stable Diffusion, all landed in 2022, and Adobe went up through that too. Adobe launched Firefly in March 2023, roughly nine to twelve months after the image boom was already obvious, which tells you it was responding to a platform shift rather than anticipating one. Investors did not care. They liked it. Firefly optimism is a substantial part of what drove the stock back to $617.78.

Adobe's identity was built on the idea that making professional visual content is difficult, slow and worth paying for. Generative video attacked all three premises at once. The company did not just look outdated. It looked redundant.

And it was caught in a genuine double bind. Keep focusing on the core business and get wiped out. Invest massively in self-disruption and get crucified by shareholders for wrecking margins. There was no path that produced applause.

To AI or Not to AI

Adobe's AI strategy is usually described as a failure of ambition or of talent. That reading is lazy. The real explanation is a single decision made once, in 2023, and everything after it follows.

Firefly is trained exclusively on Adobe Stock, openly licensed content, and public domain material where copyright has expired. That is a permanent quality ceiling against models trained on everything, and Adobe accepted it deliberately, because the point of Firefly was never to be the best model. The point was indemnification. Adobe assumes the legal risk of the output, which is the one thing a Fortune 500 marketing department cannot get from a model trained on scraped data.

You can see the same conclusion in Adobe's capital allocation. This is not a company that lacked the money for an AI push.

When management cannot find an internal use of capital that beats buying back its own stock at eleven times earnings, it is telling you what it thinks its options are. Adobe looked at the frontier model race, priced its own odds, and returned the money to shareholders.

The product decisions say the same thing even louder. Adobe now ships Sora, Runway, Pika, and Kling 3.0 directly inside the Premiere Pro timeline. It uses Google's Nano Banana model inside its own applications. It shipped its own Firefly Video model in February 2025 and by 2026 was putting its competitors' models in its own editor. A company that believed it could win on models does not architect itself as the customer of four rival model vendors.

Which brings us to the number that gets misread more than any other in this story.

AI-first ARR growth sounds enormous until you set $650M against $27.50B of total ARR. That is 2.4%. Even another year of 150% growth adds only about three and a half points to total ARR. The arithmetic simply cannot produce the kind of re-rating that AI narratives have delivered elsewhere, and the test has now been run twice: Adobe posted the Q2 numbers and the stock went to $190, then posted faster AI growth in Q3 and an after-hours bounce of about 1.3% did not hold.

The Reality: The Business Was Never the Problem

Here is the part that makes the whole story strange. Throughout a 69% collapse in the share price, Adobe was executing close to flawlessly.

Q2 FY2026 was another record: $6.62 billion, up 13%. Remaining performance obligations of $22.27 billion. Full-year guidance raised to $26.50B to $26.60B. And the market gave it nothing.

Q3 FY2026 made it eleven in a row: $6.76 billion, up 13% as reported, against a $6.69 billion consensus. Non-GAAP earnings per share came in at $6.13, up 15%, against $6.09 expected. Operating cash flow was a Q3 record at $2.52 billion. Full-year guidance went up again, to $26.576B to $26.626B of revenue and $24.45 to $24.50 of non-GAAP EPS. Both beats were narrow, about 1.0% on revenue and 0.7% on EPS.

Two things were missing, and shareholders punished the absence of both: a clear account of what Adobe becomes in the age of AI, and evidence that the account is working. Without those, no quantity of AI features would move the multiple.

What Adobe Actually Sells, and Why the Labels Mislead

Before going further it is worth being precise about the shape of the business, because Adobe's own reporting categories are more confusing than they look.

Adobe is, to a first approximation, entirely a subscription business. In Q3 FY2026 services ran at a negative 12.6% gross margin, $111M of revenue against $125M of cost, and product revenue is a rounding error at 1.0%.

The New Strategy: From How, to Why, to Whether It Landed

At some point Adobe worked out that it did not need better technology. It needed a different identity.

The old question Adobe answered was how you make an asset. That was worth an enormous amount of money for thirty years, and generative AI has been steadily reducing it to zero. The obvious next move is to answer why you make it: campaign intent, brand compliance, workflow, approvals. That is what GenStudio and the "content supply chain" pitch are.

But Adobe then discovered a third question that nobody owned at all: whether it landed. And that turns out to be where the value went.

The scale of the orchestration business is easy to underestimate. Adobe Experience Platform processes over 70 billion profile activations and 35 trillion segment evaluations per day, and delivers more than a trillion experiences a year. That is not a design company's infrastructure. That is marketing plumbing, and it produces switching costs no creative tool can match.

Adobe has a name for this. It calls it Customer Experience Orchestration, and it is claiming the category out loud: on the Q2 call, Anil Chakravarthy said Adobe leads "both the traditional marketing category and the emerging Customer Experience Orchestration category." That is the Salesforce and CRM playbook, run again.

The Board Answered the Question With a Personnel Decision

On 2 September 2026, Adobe named its next CEO, effective 1 December. Shantanu Narayen moves to executive chair after 18 years.

The board did not pick a creative software CEO. It picked a marketing platform CEO, and the creative software CEO left. If you want to know what Adobe thinks it is becoming, that decision says more than any slide deck.

Adobe pushed back on that reading at the first opportunity. Asked on the Q3 call whether Chakravarthy's digital experience background was a statement about where the board sees growth, Narayen said there had been "two great internal candidates" and that Chakravarthy "is going to own the entire company strategy." Chakravarthy pointed out that since the end of 2020 he has run enterprise field operations, representing all of Adobe to its largest customers across creativity, productivity and customer experience. Both points are fair. Neither changes which of the two candidates is still at the company.

The Dodged Bullet: Figma

Adobe paid $1 billion to walk away from Figma and spent two years being mocked for it. On the evidence, it was one of the luckiest things that ever happened to the company, and there are four independent reasons why.

The Irony: Figma Ran Adobe's Playbook Against Adobe's Disruption

Adobe tried to buy Figma because Figma looked like the modern software company Adobe needed to become. Then AI arrived and Figma had to reinvent itself on exactly the same axis.

Figma's own AI never ran on Figma's models. Its early AI features used GPT-4 and Amazon Titan. Figma Make launched on Claude 3.7 Sonnet, and now exposes Gemini, Claude and GPT through a model selector. That is precisely the conclusion Adobe reached: do not try to own intelligence, own the environment in which intelligence does useful work.

And Figma says the quiet part in its own SEC filings. Its FY2025 10-K warns that AI may reduce the difficulty of building competing products, reduce the overall value of interface design, and decrease the number of designers using Figma. Its Q2 2026 filing goes further, warning that third-party AI agents creating or modifying Figma assets could reduce the number of paid seats needed to produce the same output.

One number makes the organisational point better than any argument. In 2025 Figma generated roughly $1.06B of revenue and spent about $1.03B on R&D. That is 97% of revenue, with AI costs explicitly cited in the increase. Adobe runs R&D at 18% of revenue against a 45% operating margin target. An independent Figma can say "forget margins, we have to reinvent the company." Inside Adobe, that spend would never have survived a budget review.

Customer Asset Management: What Adobe Has Actually Been Building

Once you stop looking at Adobe as a creative software company, its acquisition history stops looking random.

Almost none of those are creative tools. Adobe has spent seventeen years buying the layer underneath the apps: analytics, commerce, marketing automation, work management, review workflows, and now search visibility.

There is a second pattern too. Adobe has stopped buying companies for their customers and started buying them for capabilities it lacks. Rilo, acquired on 2 September 2026, is a six-person team founded in 2025 that had raised $1M at a $10M valuation. Adobe bought it for a go-to-market workflow engine covering competitor intelligence, sales call analysis, and content repurposing and distribution. That is not a revenue acquisition. That is buying a competency.

Semrush: A Rounding Error That Holds Up the Roof

The Semrush acquisition is the clearest expression of the strategy, and its size and its importance point in opposite directions.

That last number is why a 1.8% acquisition is load-bearing. If product discovery moves inside the model, then everything Adobe's content supply chain produces becomes invisible at the final step, and the entire enterprise pitch collapses. Semrush is not additive revenue. It is what keeps the rest of the stack meaningful.

It is also worth being precise about what it does, because "SEO tool" undersells it. Semrush contributes outside-in demand intelligence: a database of what people are actually prompting and searching for. Adobe pairs that with the inside-out intelligence of the content sitting in Experience Manager. Aim, push, verify.

Q3 put a product name on it. Adobe Brand Visibility combines Adobe LLM Optimizer with Semrush's AI Optimization, drawing on a database of nearly 300 million real-world AI search prompts to show brands how they appear in ChatGPT, Google AI Mode, Microsoft Copilot and Perplexity. Paid customers doubled quarter over quarter. Adobe did not give a new figure for Semrush's own contribution, saying the sales motion is now the combined offering.

Shipping Inside the Competition

If you want a single fact that proves Adobe has stopped fighting for the entry point, it is this one: Adobe now ships its own agents inside ChatGPT and Claude, with Copilot and Gemini coming. The Adobe Creative Agent reaches roughly 50 creative tools across the ecosystem and is monetised through the existing credit model. CX Enterprise Coworker runs inside NVIDIA's agent platform. Acrobat ships inside ChatGPT, Claude, WhatsApp, Edge and Chrome.

Chakravarthy's own label for this is agentic software. On his first earnings call as CEO-designate he said: "I see immense opportunity for Adobe to be the leader in agentic software for creativity, productivity, and customer experience, just like we have been the leader in SaaS for those categories." What customers want, he said, is the interface of their choice, conversational or traditional, ChatGPT or Copilot, with Adobe's functionality underneath. That is the Salesforce move again: make the platform headless, so the value sits in the data and workflows rather than in whichever screen the user happens to be looking at.

Narayen's stated reason for comfort with this is worth noting, because it is the most fragile assumption Adobe made all year. Asked about partnering with Google while Google builds design tools, he said of OpenAI and Anthropic that "all of their focus right now is on code, and that's where everybody is doing a student body left on that."

That is a bet that rivals stay distracted. It sits awkwardly beside his own argument that the code playbook will run in every category, creativity included.

The Other Moat: Trust, Provenance, and the People Adobe Annoyed

There is one area where Adobe genuinely leads, and it is not a capability. It is a standard.

Across Photoshop, Lightroom, GenStudio and Journey Optimizer, Adobe embeds Content Credentials into assets: cryptographically signed, tamper-evident metadata recording the original creator, the exact tools used including which generative model, and a chronological record of every edit. Enterprise administrators can upload X.509 certificates to apply verified Brand Signatures, with invisible watermarking so an asset's origin stays provable as it moves across social platforms and ad networks.

This started as an ethics initiative and became a compliance product. The EU AI Act's transparency obligations for AI-generated visual content began biting in mid-2026, and Adobe's automated provenance means its enterprise customers comply without anyone manually tagging metadata. Adobe effectively wrote the format that regulation now requires.

Firefly Foundry is the enterprise expression of the same logic. Rather than a self-serve fine-tuning tool, Adobe assigns machine learning scientists and engineers to work directly with a client, training bespoke models exclusively on that client's approved intellectual property, guaranteed free of unlicensed material because the foundation model already is. Once a global brand has spent months encoding its visual identity into weights Adobe hosts, the switching cost is not a subscription. It is the brand itself.

The Saudi deal is the first sovereign version of that: a Firefly Foundry model built jointly with HUMAIN, a sovereign wealth fund company, tuned to Saudi visual culture and driven by Arabic prompts. Foundry has gone from "train a model on your brand" to "train a model on a nation."

Q3 added a household name on the corporate side: Disney Imagineering is integrating Firefly Foundry into its theme park design toolkit.

The Part Adobe Cannot Buy Back

None of which fixed the trust problem with individual creators.

In February 2024 a Terms of Service update triggered a serious backlash, with prominent artists claiming they could not uninstall Photoshop or cancel subscriptions without first accepting terms they believed let Adobe train generative models on their proprietary work. Adobe clarified, pointed at Firefly's licensed-only training, and the clarification was mostly accurate. It did not matter much. Combined with the early termination fees, the price rises and the cancellation obstacle course, a decade of goodwill had been spent, and a meaningful share of independent creators had already left for Affinity, DaVinci Resolve and open-source tooling specifically to get away from Adobe's commercial behaviour.

The FTC settlement makes that worse before it makes it better. A court-ordered frictionless cancellation flow removes the contractual lock-in that was quietly propping up retention, which means Adobe now has to keep individual subscribers on product value alone, for the first time in years, at exactly the moment free alternatives got good.

Butchering the Golden Goose

In June 2026 Adobe did something that almost no dominant software company ever does voluntarily. It cancelled a price increase.

Simultaneously it went aggressively freemium on Acrobat, Express and Firefly, routing intent-based search traffic straight into a working product instead of a paywall. The combined cost is roughly $500 million of annual recurring revenue, which Narayen split himself: about half from deferring the price rise, about half from the freemium push.

Three months later, asked on the Q3 call how the headwind had split between the two and when the deferred price rises would resume, management gave neither a split nor a date. Narayen said this instead:

The traffic supports the decision. Adobe.com traffic is up over 40% year over year, Business Professionals and Consumers traffic up 35%, creators up 50%.

Be careful with the headline user number. In Q3 Adobe announced a milestone of over a billion monthly active users across all its businesses, growing more than 20%. That is a sum of two very different pools. Acrobat plus Express is over 900 million, growing over 25%, with Acrobat AI Assistant users doubling in a single quarter. The creative freemium pool is a separate 100 million-plus, growing over 70%. The two grow at different speeds and monetise in completely different ways.

The precedent Narayen keeps returning to is Acrobat Reader, and the story is better than the strategy memo.

What It Is Costing

This is where the story stops being comfortable. Adobe is paying for the pivot in three places at once, and all three are visible in the numbers.

The slowdown has reached the backlog too. Remaining performance obligations slipped from $22.27 billion in Q2 to $22.16 billion, growth of 8% year over year, which Griffin Securities' Jay Vleeschhouwer noted was the first single-digit RPO growth since early fiscal 2023. Interim CFO Steve Day agreed, said the growth "is consistent with our ARR trends" and reflects the freemium push, and pointed out that RPO typically steps up in Q4 and stays roughly flat for the following three quarters. Adobe held its 10.2% full-year ARR growth target, which on its $25.66 billion starting book implies roughly $780 million of net new ARR in Q4, against about $920 million in Q4 FY2025.

Non-GAAP operating margin fell again, to 44.0% in Q3, the lowest in the published eleven-quarter series, and Q4 is guided at about 44.0% as well. Non-GAAP operating expenses grew 16.2% year over year against 12.9% revenue growth. Compute is not free, and it is showing up in the actuals rather than in analyst projections.

And a claim you will hear constantly, that Adobe is getting leaner, is simply not true.

Headcount is 34,206, a dip of 122 from Q2 but 8.5% above the 31,520 of a year earlier. The 30,709 figure that circulates in commentary is from Q4 FY2024 and is nearly two years stale. The only genuine trim anywhere in Adobe's disclosures is Advertising Cloud, which was scaled down for margin and then written down by $70 million in Q2 FY2026.

The Bridge

While all this plays out, one mechanism is holding the stock up, and it is not a product.

Adobe has retired 13.4% of its own diluted shares in ten quarters, about 9.5 million of them in Q3 alone. It authorised a fresh $25 billion buyback in April 2026 on top of a previous authorisation with only $2 billion left, which Narayen said would have been completed in under eleven quarters. Exiting Q3 the old authorisation is fully used and about $24.55 billion of the new one remains. Against a market cap near $108 billion in late August, that is nearly a quarter of the company. Shrink the denominator fast enough and earnings per share grows whether or not the strategy works.

Bear and Bull, Stated Plainly

The dispersion is the story. Out of 27 tracked analysts, 15 say hold, 8 say buy, 4 say sell. The average target sits around $271 against a price near $275, implying nothing at all, while the individual targets run from $190 to $487. That is not a consensus with error bars. That is a complete breakdown in agreement about what this company is.

So What Is Adobe Now?

Adobe is winning the enterprise and losing the creator. It has traded tool dominance for workflow, compliance and data gravity at the top of the market, while Canva burns down the bottom of the funnel and Figma owns a stage Adobe cannot reach.

The strategy is coherent, and it is not an AI strategy. It is a distribution strategy wearing AI vocabulary, because the market only asks AI questions. Every concrete decision Adobe made in 2026 was about arrival, not intelligence: freemium routing, deferred consumer pricing, Semrush, agents shipped into rivals' chat windows, and a $4 billion commitment to give 27 million Saudi citizens a year of free Firefly and Express with a state-backed model tuned to Saudi visual culture.

Investors do not want to see the old Adobe generate slightly more revenue with AI attached. They want to know what the new Adobe is. In September 2026 they finally got an answer, in the form of a CEO appointment, and they marked the stock down 7% on it. Six days later Adobe posted a record quarter, raised guidance and reported AI-first ARR growing more than 150%, and an after-hours rise of about 1.3% did not hold.

Which leaves a bet with an actual scoreboard. Adobe gave up half a billion dollars of recurring revenue to buy habit it intends to monetise from FY2027, and Narayen dated the payoff himself: it "will play out, I think, over 2027." If that conversion appears, this was the year Adobe stopped defending a shrinking castle and started building a bigger one. If it does not, this was the year Adobe's pricing power ended and freemium was the cover story.