Adobe

Adobe makes the tools that professional creative work is made with — Photoshop, Illustrator, Premiere, InDesign — and it invented the PDF. If you have ever edited a photograph, opened a document that looked identical on every machine, or seen a magazine, an advertisement or a film title, you have used its output. Photoshop is one of the few products in history to have become a verb.

It is also, on the numbers, an exceptional business: an 89% gross margin, a 36% operating margin, a 36% return on invested capital, and free cash flow of $26.44 per share against earnings of $17.98 — a company that converts profit into more cash than it reports. Since 2021 its revenue has grown 60%, from $15.8 billion to $25.2 billion, and its earnings per share have grown 74%.

Over that same period, the shares have fallen 64% — from a peak of $688 in November 2021 to $250 today. The market capitalisation has slipped below $100 billion. Almost none of that fall is earnings; virtually all of it is a re-rating. Adobe traded at 40 to 60 times earnings for most of the last decade and trades at 14 times trailing earnings and roughly 10 times next year's today — a multiple compression of about 78%.

The reason is a single argument, and it is worth stating in its strongest form: generative artificial intelligence may make Adobe's tools unnecessary. If a marketing manager can describe an image and receive it, she never opens Photoshop, never learns it, never pays for it. Switching costs protect you from a competitor doing the same job differently. They do not protect you when the job itself stops being done. The market has sorted software into companies AI will enrich and companies AI will destroy, and put Adobe firmly in the second pile. This report asks whether the accounts agree.

"AI is accelerating customer behavior at an unprecedented speed." — Shantanu Narayen, Adobe's chief executive of eighteen years, June 2026 — three months after announcing he would step down

We have published two reports that bracket this question. Salesforce was a wonderful franchise cheap on AI fear, where the numbers had not broken — we called it Buy the Fear. PayPal was a moat genuinely eroding, where the numbers had broken — share loss, take-rate compression, margin decline visible in the accounts. The single most useful thing this report can establish is which of those two Adobe is. The answer, as you will see, is closer to Salesforce than to PayPal — with four qualifications that are real and that most write-ups skip entirely.

Two things from this history matter for the investment case today. First, Adobe has faced an existential platform question before and answered it correctly — the 2013 subscription conversion was deeply unpopular and enormously value-creating, and the management team that ran it is largely the one running the company now. Second, the PDF decision is the template management is explicitly reaching for again: give away the commodity layer, own the workflow above it. In Part VI you will see Adobe doing exactly that with AI models — integrating rivals' engines into its own products rather than trying to beat them. Whether that is strategic genius repeating itself or a company conceding it cannot win on model quality is the question the market is currently answering in the negative.

Buffett avoided technology for decades not because he thought it was bad business, but because he could not say what a technology company would look like in ten years. Adobe in 2026 is the purest example of that problem we have put in front of you. The economics are magnificent and legible; the durability is genuinely contested by serious people on both sides. The honest position is not "buy the fear" or "avoid the melting ice cube" — it is that this requires a judgement about the future of a technology, and if you cannot make that judgement confidently, the correct position size is small no matter how cheap the multiple. We will return to that in the letter, because it is more useful than pretending to certainty.

The inversion in those two lines is the most interesting non-obvious fact in the company. The conventional story is that Adobe's professional core is safe behind switching costs while the cheap end gets eaten by Canva and free AI. The actual numbers say the opposite is happening: the consumer and prosumer group is growing at 15–16% and the professional creative core at 11–12%. The part everyone assumes is most vulnerable is currently doing the heavy lifting, and the part everyone assumes is protected is growing more slowly. That may be because Acrobat is a genuinely under-appreciated annuity — PDF is unavoidable in business — or it may be that professional creative seats are already feeling pressure. Both readings fit the data, and Adobe's new disclosure format makes it harder to tell which.

The bull framing, stated fairly. Adobe's moat is not the software — it is the industry's interoperability convention. A designer does not use Photoshop because it is the best pixel editor; she uses it because the agency expects a .psd, the printer expects a .indd, the editor expects a .prproj, the university taught it, and every colleague on the project has a decade of muscle memory in it. That is a network effect wearing the costume of a switching cost, and it compounds with every asset created and every handoff made. Layer on Adobe Fonts, colour management, and the fact that Adobe gave PDF away as an open standard in 2008 and kept Acrobat — making the format universal while owning the professional workflow around it — and you have one of the most durable positions in software.

The bear framing, and it is the better argument. Switching costs protect you from a competitor doing the same job differently. They do not protect you from the job becoming unnecessary. If a marketing manager describes an image and receives it, she never enters the workflow the moat is built around — the wall stands, but the traffic routes past the city. This is precisely the disruption pattern Buffett has always said he cannot underwrite, and it is why the multiple has halved twice.

★ And there is a second attack that has nothing to do with artificial intelligence at all, which most coverage folds into the AI story and should not. On 29 October 2025, Canva took Affinity — the professional Photoshop/Illustrator/InDesign substitute it had bought for about $380 million — merged its three applications into one, and made it free, permanently, for everyone. Canva's own words: "There's no catch, no stripped-back version, and no gotchas." It opens Adobe's own file formats natively. A million people signed up in four days. Canva can afford this because it earns roughly $4 billion of recurring revenue from 265 million monthly users and has been profitable for eight years; it monetises by charging for AI features inside the free product. Set that against Adobe raising Creative Cloud prices roughly 11% in the same twelve months. One company took a professional creative suite to zero while the other raised prices — and Adobe's single-app subscription now costs about $276 a year against a free competitor that reads its files. That is a direct assault on pricing power, and it is not a generative-AI story.

What the evidence actually shows. Two things cut in Adobe's favour and one cuts hard against. In favour: gross margin is at an all-time high of 89%, up from 87.7% in 2022 — a company losing a moat does not usually expand its gross margin; and revenue growth has never broken, running around 10–11% throughout the AI era. Against: Adobe raised Creative Cloud pricing about 11% in mid-2025, cut the downgrade tier's AI credits from 1,000 to 25, took real customer backlash — and then deferred further planned price increases in FY2026. That is the single most concrete evidence in the file of pricing power under strain, and it is management's own decision, not an analyst's inference.

I score the moat a 7: genuinely deep at the professional tier, where format lock-in and workflow interoperability remain formidable and where no AI tool yet does the job — marked down because the moat is being attacked at the entry point rather than the centre. Adobe is not losing art directors. It may be losing the next generation of them before they ever become art directors, which is a slower and more dangerous problem.

We promised a test that settles this better than opinion, and there is one: is Adobe's recurring revenue growth accelerating or decelerating? If AI were driving more people into Adobe's tools, it would show up as accelerating annual recurring revenue. Here is the honest answer.

The verdict on the test: decelerating — clearly, steadily, and on management's own guidance. Note the trap in that table: the headline +12.5% in the most recent quarter is flattered by the Semrush acquisition, which brought roughly $480 million of bought recurring revenue. Strip it out and organic growth is below the prior year's exit rate. Anyone quoting the 12.5% without that adjustment is repeating a number that is technically true and analytically misleading.

Now the nuance that stops this being a simple bear point, and it is important. Management chose part of this deceleration. On the June 2026 earnings call, Adobe lowered its second-half expectations and explained why: roughly half the shortfall came from deferring planned price increases, and roughly half from going, in the chief executive's words, "full steam" on freemium — deliberately giving product away to build a funnel. That is a management team choosing users over near-term revenue.

There are exactly two readings of that, and intellectual honesty requires holding both. The generous reading is that this is the correct response to an AI-driven land grab: get 90 million free creative users inside your product now, monetise later, exactly as Adobe made PDF universal and then sold Acrobat. The unkind reading is that "strategic freemium" is the polite way to describe losing pricing power — you do not defer price increases from a position of strength. The honest tell to watch is conversion: if free users keep compounding and paid revenue keeps decelerating for another four quarters, it was pricing power loss. If paid growth stabilises while users compound, it was strategy.

And there is a structural point in Adobe's favour that deserves more weight than it gets. Adobe has integrated rivals' AI models — OpenAI's, Google's and others — directly into its own products, charging for their use through its credit system. That is the PDF playbook again: commoditise the layer below you, own the workflow above it. If the model itself becomes a commodity — and every price trend suggests it will — then the durable value accrues to whoever owns the interface, the file formats and the professional workflow. Read charitably, this is Adobe positioning itself as the toll booth regardless of who wins the model war. Read unkindly, it is an admission that Adobe cannot win on model quality. We think the charitable reading is more likely correct, and we hold it with less confidence than we would like.

★ Two events in 2026 cut hard against the simple "free AI eats Adobe" thesis, and almost nobody has connected them. First: OpenAI shut Sora down. The most-hyped consumer video-generation product on Earth — launched to enormous fanfare in 2024, relaunched as a social app in 2025, with a billion-dollar Disney licensing deal attached — was announced for closure in March 2026 and the app shut in April, reportedly costing around a million dollars a day to run. The Disney money never changed hands. Second: Stability AI, the company that started the open-source image-generation wave, has not shipped a flagship image model since October 2024 and has pivoted to audio and enterprise services. The lesson in both is the same and it is the most underrated fact in this debate: the free thing is extraordinarily expensive to run. Generation at scale is a compute bill, and compute bills require either a subscription business or a very patient investor. Adobe has 850 million monthly users and $9.85 billion of free cash flow to pay for inference. Meanwhile the surviving model layer has become a supplier: the same Black Forest Labs models sit inside Adobe, Canva, Snap, Meta and xAI, and Google will sell you an image for about four cents. When everyone rents the same engine at the same price, the differentiation moves back up to workflow, file formats, colour management, rights and enterprise procurement — which is precisely the ground Adobe still owns. Adobe put Google's newest image model inside Photoshop the day it launched. That is not the behaviour of a company being outrun.

In September 2022 Adobe agreed to buy Figma, the collaborative design tool, for $20 billion — roughly fifty times Figma's recurring revenue, and the largest software acquisition of its kind. UK and EU regulators opposed it. In December 2023 Adobe abandoned the deal and paid a $1 billion termination fee for no assets, no people and no intellectual property.

Both halves of this deserve to be told. Adobe got lucky. Figma has more than tripled its business and is worth roughly a third less than Adobe offered — the regulators saved Adobe's shareholders from what would today be a paper loss of several billion dollars. That is genuinely true, though it is a snapshot taken in a software bear market that has hit Adobe's own multiple just as hard.

But Adobe still paid a billion dollars in cash for absolutely nothing, and the process revealed something more damaging than the fee. In examining the deal, the UK regulator found that Adobe's own competing product, Adobe XD, held roughly 5–10% of the market against Figma's 80%-plus. Adobe tried to compete in collaborative design and lost badly; then tried to buy the winner and was blocked. That is the most revealing disclosure in the entire saga, and it belongs in any honest assessment of how well this company defends contested ground.

The sting is in the present tense. Figma is now a listed rival expanding directly into Adobe's territory — it launched Figma Draw (aimed at Illustrator), Figma Buzz (aimed at Express), Figma Slides and Figma Sites, going from four products to eight in a single conference. And its revenue growth is accelerating — 38%, then 40%, then 46% — while Adobe's decelerates. Both stocks are down heavily, because the market is repricing all design software for AI risk. But if you ask whether Adobe's competitive position in design has improved since 2022, the answer in the numbers is no.

Two of the three most senior roles at Adobe turned over inside three months, in the middle of the most consequential strategic decision in the company's life. Adobe is deciding how much of its franchise to give away free in order to defend it against AI — and it is deciding this with a chief executive who has announced his departure and a finance chief who is temporary. The board has a special committee running the search, targeting a successor in place for the FY2027 planning cycle.

The unkind reading is rats and a ship. The fairer one is that Narayen, at the natural end of a very long and successful tenure, is handing over a franchise that plausibly needs a different kind of leader for a different era. We lean to the fairer reading. But it is a real risk and it compounds the others: this is a company changing its disclosure format, changing its pricing strategy, and changing its leadership, all at once, while its critics say its moat is melting.

Two further observations cut against the bulls. The first is that shareholders are already unhappy: at the April 2026 annual meeting Adobe's advisory vote on executive pay passed with roughly 50.7% support — 148.8 million votes for against 145.0 million against, effectively a coin flip, for a company paying its chief executive over $51 million. A near-failed say-on-pay is a rare and unambiguous signal of owner discontent, and it sits oddly beside insider ownership of well under 1% (all 18 directors and executive officers together hold about 0.2% of the company). Nobody running Adobe owns much of it. The second is that Adobe is textbook activist bait — a hundred-billion-dollar franchise at ten times free cash flow, with 89% gross margins, no dividend, and a leadership transition under way. Yet no activist has appeared, and insider ownership is essentially nil (0.84%, with the CEO holding under half a million shares). Either the smart money believes the moat really is melting, or the opportunity is not yet obvious to it. We note it without resolving it. I score management a 6: a superb operating and buyback record, weighed against the Figma billion, the disclosure change, and a leadership vacuum at the worst possible moment.

Read this table and then read the share price again. Revenue quadrupled over the decade. Gross margin is at an all-time high. Operating margin expanded five points in the most recent year. Earnings per share rose 35%. Free cash flow exceeds reported earnings. Return on invested capital is 36%. The balance sheet is essentially unlevered and the financial-health scores are near the top of our board. There is nothing in these accounts that looks like a business losing its moat — which is precisely the point, and precisely why this is difficult: the bear case is entirely about the future, and the future does not appear in the accounts until it is too late to buy cheaply.

Two honest deductions. First, stock-based compensation runs 8.1% of revenue — $1.94 billion a year, a genuine cost to owners that the headline free-cash-flow yield of 10.7% ignores. Adjust for it and the yield is closer to 8.3%, which is still excellent but no longer extraordinary. Second, the one line that does point the wrong way is the one that matters most to the debate: recurring revenue growth is decelerating, from 11.5% to a guided 10.2%. Everything else in this table is a company performing well. That single line is the entire bear case with numbers attached — and, as Part VI showed, roughly half of it is management's own deliberate choice.

Start with the comparison that should be in every write-up of this company and is in almost none. Adobe trades at a discount to Salesforce on every single multiple — while earning an operating margin fifteen points higher and a gross margin near 90%. Both are down about 31% over twelve months. This is not an Adobe-specific verdict; the market has sorted application software into "AI winners" and "AI roadkill," and Microsoft — seen as a winner — trades at roughly 2.4 times Adobe's forward multiple on a 1.9% free-cash-flow yield. Adobe is being priced as a category, not as a company.

What you are actually being offered: a business with an 89% gross margin, a 36% return on invested capital, no meaningful debt, free cash flow that exceeds reported earnings, and a buyback retiring roughly 6% of the shares a year — at ten times next year's earnings and about nine times free cash flow. On the consensus path to FY2028 you are paying eight times. For that to be the right price, Adobe's earnings must not merely stop growing; they must decline materially and permanently. Nothing in the accounts yet suggests that, and the discounted-cash-flow model — a blunt instrument, but a useful sanity check — puts fair value at $335, some 34% above the price.

And now the discipline. Cheapness is not an argument if earning power is falling; that was the lesson of PayPal. The four things that stop us being emphatic are these: growth is decelerating; management did defer planned price increases, which is what losing pricing power looks like from the inside; the company did stop disclosing the metric its critics were watching; and the chief executive and finance chief both left within three months. None is fatal. Together they are the difference between "cheap and fine" and "cheap and genuinely uncertain," and they are why our verdict is constructive rather than emphatic.

So: "Priced for a Melt the Numbers Don't Show." I score valuation a 9 — the most attractive on our board — and would accumulate here, adding toward $200, near the $190 low this stock touched within the past year, where the free-cash-flow yield exceeds 13% and you are paid handsomely for a future nobody can forecast. But size it as what it is: a genuinely cheap franchise whose ten-year picture is unknowable, which is an argument for a smaller position at a better price, not a bigger one.