If you sat through enough earnings calls this year, you started to notice a pattern. Sooner or later some analyst would clear their throat and ask the question every executive has learned to dread the one lurking under every tech valuation in the age of ChatGPT. Won't AI just make your company obsolete? What's striking isn't that CEOs had an answer ready. It's that so many of them, across wildly different businesses, gave almost exactly the same one.
The question everyone was really asking
Behind the analyst politeness, Wall Street's worry is blunt. If anyone can rent a large language model that writes, codes, plans and chats, what stops that model from eating the companies caught in the middle? Why book through a travel site, browse a review app, or pay for sales software when an AI could just… do it for you? Put that to the bosses of Airbnb, HubSpot, Yelp, Pinterest and Roku and you don't get a solo. You get a chorus.
"The model isn't the moat"
Their shared answer runs like this: the AI model is a commodity the valuable, defensible thing sits on top of it.
Airbnb's Brian Chesky has been the bluntest. The underlying tech, he argues, "is not proprietary" it's available to every app on Earth through an API. In his telling that's freeing, not threatening: pretty soon, he says, every company becomes an AI platform if it makes the shift.
HubSpot's Yamini Rangan drew the sharpest line between AI that impresses and AI that pays the gap between the tool that produces words and the one that wins deals. Output versus outcome.
Yelp, Pinterest and Roku sang variations of the same note. A chatbot can generate a fluent paragraph, but it can't manufacture two decades of real restaurant reviews, a live map of what millions of people actually pin and buy, or first-party data on what you watch.
Why it's a genuinely good argument
Strip away the earnings-call polish and there's a real idea underneath. If everyone can rent the same model, the model can't be what sets you apart by definition. What a rented LLM can't copy is the stuff these companies spent years accumulating: verified identities, mountains of reviews, customer histories, and the trust that makes someone hand over a credit card.
"Data is the moat" has been a cliché for a decade. The LLM era gives it teeth, because it draws a hard line between the part of the stack that's now a commodity — the raw intelligence and the part that isn't: the proprietary data and relationships you feed it. That's the real weight behind Chesky's line. If the model is just a layer you bolt on, then the winners aren't the ones with the cleverest AI. They're the ones sitting on data a model can't scrape.
THE HONEST CATCH
Now the part the earnings calls skip. These are chief executives defending their share price to Wall Street of course they say AI won't kill them. It's the most predictable message in the world, and it should be read as an argument, not a verdict. The bear case is real: AI "agents" that book, buy and decide on your behalf could quietly disintermediate the very apps claiming to be safe if an assistant books your trip, does it matter whose logo was on the listing? A data moat can be eroded by scraping, partnerships or synthetic data. And the same logic that reassures Airbnb is a death sentence for the thin "wrapper" apps whose only asset is a model everyone else can also rent. Worth noting, too: this is an evergreen pattern pulled from months of earnings calls a way of thinking, not a fresh headline.
What to watch
Agents versus incumbents. The real test of the "data moat" isn't chatbots it's autonomous agents that transact for you. Watch whether they route through these companies or around them.
Output versus outcome. Rangan's distinction is the one to keep. Plenty of AI will produce impressive words; far less will reliably win the deal, book the stay, or close the sale. The gap is where the money is.
The wrapper shakeout. If the moat really is proprietary data, the thin apps with none of it are the first dominoes. Watch which "AI startups" quietly disappear.
Whose data actually compounds. Not all first-party data is a moat. Watch who turns theirs into something a rival genuinely can't replicate and who just says they have.
EDITOR'S TAKE
The useful move here isn't to believe the CEOs it's to borrow their framework. Strip out the stock-pumping and you're left with a genuinely clarifying question for almost any business, including a one-person one: in a world where anyone can rent the same intelligence, what do you own that a chatbot can't copy? The verified relationships, the proprietary data, the trust earned over years that's the moat. The model was never going to be it. Audit what a rented AI couldn't reproduce about what you do, and you're looking at the part actually worth defending.
One frontier-tech story, in plain English, every weekday.
Frequently asked questions
Why do CEOs say their data, not the model, is the moat against AI?
Because the same large language models are available for any company to rent, so the model alone can't set you apart. What can is the proprietary data a chatbot can't copy verified identities, years of reviews, customer histories, trusted relationships. As Airbnb's Brian Chesky put it, the models "are not proprietary," and every company can become an AI platform if it makes the shift.
Does that mean these companies are safe from AI?
Not necessarily. These are CEOs reassuring Wall Street, so treat it as an argument, not a guarantee. AI agents that transact on your behalf could disintermediate even data-rich incumbents, and a data moat can erode over time. The claim is strongest for companies whose data is genuinely hard to replicate, and weakest for thin apps whose only asset is a rentable model.
What's the takeaway for my own business?
Audit what you own that an LLM can't reproduce: proprietary data, verified relationships, and trust built over time. In an era where the intelligence itself is a commodity, that's your defensible layer not the model you plug in.
Sources
