Every chief executive in America has spent two years saying artificial intelligence is transforming their company. Almost none can point to a line in the financials where it shows up. After Airbnb (ABNB) reported second-quarter results, Brian Chesky went further than the usual formulation: "I think now it's safe to say AI is the best thing to have happened to Airbnb." The stock rose roughly 15% the next day.
He said it in a CNBC interview published Aug. 7, following the company's earnings call the evening before. In the same interview, he said Airbnb will spend "a lot more" on AI tokens this year than it originally forecast, a mid-year increase in a cost line, disclosed on television rather than in guidance. He attached no dollar figure to it, and nobody should invent one on his behalf.
More Top Stocks Daily: Go behind Wall Street’s hottest headlines with Barchart’s Active Investor newsletter.
The reported numbers are the strongest part of his case because they are the part anyone can check against a filing. Airbnb said revenue rose 17% year-over-year (YoY) to about $3.6 billion, gross booking value rose 16% to roughly $27.2 billion, and net income landed near $816 million. Adjusted EBITDA of about $1.3 billion worked out to a 35% margin, and the company guided full-year margin up to at least 35.5%. That last item is the closest thing to hard evidence here: a company raising AI spending and raising margin guidance in the same breath, rather than asking investors to wait.
Chesky's Justification
His justification is worth reading slowly. The cost of inference, Chesky argued, "pales in comparison" to what Airbnb earns on each booking and to the revenue that comes from building products faster. Inference is what you pay every time a model actually runs, every support reply drafted, every listing summarized, and every search reranked. It is a per-use cost rather than a one-time build, which is precisely why investors have been nervous about it. His claim is that on Airbnb's unit economics, the ratio is not close.
The evidence then arrives in two very different tiers, and most coverage has blended them. Tier one is the reported financials: revenue, bookings, net income, margin, and guidance. Those sit in a shareholder letter, and anyone who wants to argue can read it against what analysts expected going into the quarter.
Tier two exists only because Airbnb says so. The company told investors AI has cut development time for some initiatives by as much as 60% while lifting features and improvements shipped in the first half by nearly 80% year over year, noting both hedges, "some initiatives" and "as much as," which several outlets dropped on the way to reporting a flat 60% cut. Airbnb also said roughly 60% of its new code is AI-coauthored, about twice the industry average, without naming whose estimate that average is. And it said its AI support assistant works in 50-plus languages and resolves about 45% of inquiries without a human, with support cost per booking down around 16%.
A Closer Look Raises Questions
Nobody outside the company can audit any of it. The support figure deserves particular care because "resolved" is a defined term in support telemetry, and in most systems, a session that ends because the customer gave up scores identically to one that ended because the problem was “resolved.” That does not make the number false. It makes it a number Airbnb produced, measured by Airbnb, using Airbnb's definitions, a different category of fact from a filed income statement.
Airbnb also undercut the AI-did-it reading itself by naming two other drivers on the same call. Reserve Now, Pay Later accounted for more than 20% of gross booking value in the quarter, and hotels are growing roughly three times faster than the core homes business. Both are ordinary product and category expansions. Crediting the whole quarter to AI overstates the case on the company's own account.
What generalizes is narrower than the headlines suggest. If inference really does pale next to incremental revenue, that weakens the AI-capex bear case for asset-light platforms specifically, the companies buying tokens, not the ones building data centers. It says nothing about the firms selling the GPUs, whose economics run the other way. Readers wanting to test which other businesses show AI in the margin line rather than the narrative can start by screening for companies raising guidance and spending at once and by checking Wall Street's rating on the stock before assuming this is priced.
Chesky made this argument on a day his stock rose 15%, which is the day a chief executive is most inclined to make it. One quarter is not durable operating leverage, and 50 basis points of margin guidance is a real number but a small one. He raised his AI budget mid-year without giving a figure. Next quarter is where that either shows up in the margin or stops being an argument.
On the date of publication, Caleb Naysmith did not have (either directly or indirectly) positions in any of the securities mentioned in this article. All information and data in this article is solely for informational purposes. For more information please view the Barchart Disclosure Policy here.
More news from Barchart
Etsy's Workforce Reduction Sends a Massive Signal on Its Growth Strategy. How to Play ETSY Stock Now. Nvidia AI Chips Are Now an Asset Class. BlackRock CEO Larry Fink Just Compared Them to Mortgage-Backed Securities. Airbnb CEO Brian Chesky Says AI Is 'The Best Thing to Have Happened' to His Company — And That the Millions Spent 'Pales In Comparison' to What It Brings Back Tesla Stock Surged 40% From Here Last Year. Here's How to Play TSLA Stock in 2026.