What $21 billion taught them about execution, ownership, and the one thing money couldn’t buy.
In September 2022, Adobe agreed to pay $20 billion for a company with a fraction of its revenue. Figma wasn’t a feature Adobe was missing — Adobe already owned design software. What Adobe couldn’t produce with its own execution machine was the category Figma had built from nothing: real-time, browser-based, collaborative design. Regulators in the US, UK, and EU agreed that was exactly the problem, and blocked the deal on those grounds. It died in December 2023. Adobe paid a $1 billion breakup fee and walked away with nothing to show for it.
Figma went public in July 2025 at roughly $19 billion. By the end of its first trading day, it was worth close to $68 billion — more than triple what Adobe had offered three years earlier.
Adobe had every execution advantage that exists in enterprise software: distribution, installed base, sales infrastructure, balance sheet, brand. None of it let them replicate what Figma had decided to build. So Adobe tried to buy the decision instead of making it. And still couldn’t close it.
Adobe hit that wall with a $20 billion checkbook. It isn’t the only place that wall shows up.
The pattern doesn’t stay in software. Rhode launched with three products, no retail partner, and no incubator — e.l.f. Beauty bought it for up to $1 billion three years later, the fastest run to that number in beauty history, because Hailey Bieber owned a point of view no competitor could license. Olipop owns 60% of the category it invented and keeps compounding. Liquid Death didn’t own a category — just marketing — and growth slowed from triple digits to 27% the moment every retailer had its own edgy canned water; a late-2024 share sale marked its valuation down by a third. Olaplex didn’t own one either: once L’Oréal and Redken matched its bond-repair claims, its stock fell more than 90% before Henkel bought what was left. Poppi didn’t wait to find out which side it was on — PepsiCo paid $1.95 billion for it while it was still winning. Software, fashion, beauty, beverage: same test, every time. Own the category and someone pays a premium for it. Own only the execution and someone eventually buys you cheap, or doesn’t bother.
AI is about to put that gap in front of every company that never had a checkbook, a factory, or a shelf slot to compete with — the only thing that changes is the price of finding out. It’s about to become the normal condition of every market, because AI does to every category what it’s already doing to marketing: make execution instant, cheap, and identical across every competitor at once.
McKinsey’s growth research found that only 7% of corporate growth comes from taking market share — the rest is driven by “where to play” choices.[1] Not my phrase. Theirs.
That 7% built an entire consulting industry on its own. Optimize the funnel. Maximize reach. Drive penetration. Win the share battle inside a category someone else already defined.
None of that is wrong. It’s just not strategy. It’s operations.
There’s a reason that 7% is the part AI absorbs first. The entry-point-finding, asset-testing, media-optimizing work is the most automatable kind of growth work there is, because it was always built to be repeatable — run the same play twice, benchmark it, optimize it. That’s precisely what made it programmable.
And AI is about to make the difference impossible to hide. 87% of marketers now run generative AI weekly, up from 51% two years ago. It already finds the entry points. Tests the assets. Generates the creative. Optimizes the media buy in real time — this is what Performance Max and Advantage+ already do by default, not as an edge case.
Ask what “AI and marketing” means today and you’ll hear the same handful of answers on every panel: visibility and discoverability — will the algorithm even mention you. AI-generated ads — faster creative, more variants, cheaper media. The idea that a brand now has two customers, a human one and a machine one, so you optimize for both. The collapse of the customer journey. The share of shoppers who ask a chatbot before they buy. Every one of those is still penetration, wearing a new interface. Visibility and discoverability are entry-point-finding pointed at a new engine. AI-generated ads are asset-testing with a faster tool. “The machine is a customer too” is penetration logic aimed at a second gatekeeper instead of the first. None of it asks what should exist for either customer to find. It only asks how to be found faster inside whatever already exists.
That’s not just faster execution. It’s commoditization, and the mechanism is worth naming: a capability commoditizes when it’s automatable, replicable without proprietary skill, and available to every competitor through the same infrastructure at the same time. Penetration clears all three now. Google and Meta don’t sell you a faster version of the loop — they run the same loop, on the same terms, for every advertiser simultaneously. When the skill that used to separate a good media team from a mediocre one becomes a utility everyone taps identically, the edge that lived there goes to zero. Nobody out-executes a rival running the identical autonomous agent against the identical inventory.
The value doesn’t vanish. It relocates — to whoever owns the utility. Every dollar penetration used to pay an agency or a brand team for craft now pays the platform for access. You can watch the toll booth move in the headcount data: 23% of agencies cut junior copywriters last year. 31% plan more cuts this year. Autonomous agents running full campaigns more than doubled in two quarters. That’s not efficiency. That’s margin changing hands.
What’s commoditized is the craft — the testing, targeting, and iteration loop. What isn’t, yet, is the raw material the loop runs on: proprietary data, distribution position, demand already earned. Two competitors running the identical agent still get different results if one of them has something worth optimizing against. Adobe proved this before AI ever touched the category — it had the execution and still didn’t have the $68 billion. What it didn’t have was the category itself.
McKinsey’s research on what it calls “arenas” — the industries that broke away from the pack — makes the same case at a larger scale.[2] From 2005 to 2020, arena revenues grew 10% a year against 4% for everyone else, and the arenas nearly tripled their share of total market cap. The clearest of those shifts came from categories created by Alphabet, Amazon, Meta, and Microsoft — cloud, consumer internet, e-commerce. None of them won by taking share inside a category that already existed. They won by finding demand that didn’t have a category name yet, then building the category around it. Cloud computing wasn’t on an entry-point map in 2005, because entry-point maps only show you categories that already exist. A demand space isn’t a segment inside a category. It’s a pattern of unmet need that hasn’t been assigned a category yet — and AI can search inside a category’s entry points instantly, but it cannot search for a category that doesn’t exist, because there is nothing yet to search.
This is also where the money actually accrues, and the difference is ownership. Penetration was never something you owned — it was rented, one auction at a time, from whoever runs the platform. Stop paying and the reach disappears the same quarter; the media budget that won last quarter buys the same auction this quarter, no better than the year before. Discovery is the only kind of growth a company actually owns. Demand discovered early compounds into share of a category, share compounds into pricing power, pricing power compounds into economic profit that outlasts any single campaign — and once a category exists in the market’s head, nobody else is bidding for that slot. That’s category creation’s real advantage over penetration: not that it’s nobler, but that it’s the only growth you keep after you stop paying for it.
So: 75% of marketers have no AI roadmap. Only 49% can measure what their AI spend is even doing. They’ve automated the 7% and left the rest unmanaged.
When execution is free, the only thing left to compete on is the choice AI can’t make for you — what demand to create, which category to own, what you refuse to become.
Reach, availability, and penetration aren’t going away. They’re becoming the operating system everyone gets by default.
Penetration is a dial you turn. Except now the platform turns it for you. You never owned that dial — you were only ever renting time on it. The only thing still in human hands is which category it gets pointed at, and that’s the only growth anyone gets to keep.
Figma didn’t win the design-software category by out-executing Adobe. It decided the category should exist before Adobe knew there was a decision to make. Neither did Rhode, and neither did Olipop. Adobe had every execution advantage there was, and still had to try to buy the decision instead of making it.
That was always the job.
Sources
[1] McKinsey & Company, “Now, New, Next: How Growth Champions Create New Value.” mckinsey.com
[2] McKinsey Global Institute, “The Next Big Arenas of Competition,” October 2024. mckinsey.com