The 2025 results have the rare distinction of being both excellent and ignored. Revenue rose 13% to $26.9 billion. Adjusted EBITDA jumped 20% to $9.9 billion, lifting the margin to 36.9%. That marks the third straight year of expansion, after 33% in 2023 and 35% in 2024. Adjusted EPS hit $228. Over the fiscal year, Booking returned $8.2 billion to shareholders, more than its GAAP net income.
Q4 deserves a closer look. Revenue up 16%, room nights up 9%. The United States, long a laggard, tipped into the low double digits. On February 19, 2026, the day the numbers dropped, the stock lost more than 8%.
The invisible toll
The most structurally significant development of the year is also the least talked about. In twelve months, the share of bookings where Booking directly processes the traveler's payment climbed from 63% to 70%, some $130 billion run through merchant mode. Booking does not charge its hotels a higher commission. But by handling payments in over 50 currencies, it captures incremental revenue on each transaction that the annual report confirms exceeds the associated costs. It is a structural improvement in profitability, hiding in plain sight.
The strategic implications run deeper still. The day a hotelier in Bali depends on Booking to process a Swedish tourist's Visa card in kronor, remit the proceeds in rupiah two weeks later, and handle a potential dispute at three in the morning, switching platforms stops being a pricing decision. It becomes an infrastructure decision. And infrastructure, unlike algorithms, does not get swapped out with a click.
What everyone is talking about
Now, let's talk about what everyone is talking about.
Generative AI is going to disintermediate online travel agencies. That is the thesis. It is taken seriously enough that a stock beating consensus quarter after quarter trades at a lower multiple than a toothpaste company. Colgate-Palmolive, to name one, trades at roughly 26 times 2026 earnings on 4 to 5% organic growth. Booking, at 17 times, with 15% earnings growth. There is something almost poetic about the idea that toothpaste has become a safe haven from artificial intelligence.
The thesis deserves more than a shrug. An AI assistant capable of planning a trip end to end would genuinely change the game. The question is: for which business, exactly?
Booking Holdings houses two businesses the market treats as one. KAYAK is a metasearch engine, a smart billboard that redirects the traveler to another site. That is precisely the kind of service an LLM can replace, and Booking acknowledged as much by writing down KAYAK by $457 million this year. The company buried the metasearch business in its books. The market, for its part, keeps burying the stock.
Yet KAYAK and its associated advertising revenue account for roughly 4% of the group's total revenue. Four percent. The remaining 96% are full-service booking platforms (Booking.com, Priceline, Agoda) that process payments, guarantee reservations, handle cancellations, and provide after-sales service across 195 countries. An LLM can suggest a hotel. Processing a Visa card in Swedish kronor, remitting the amount in rupiah, handling a cross-border dispute: that is a different business altogether.
The most telling precedent remains Google. Twenty years in travel, a dominant position in search, billions invested, and still no booking agency. Booking, meanwhile, has already signed partnerships with ChatGPT, Gemini, Microsoft, and Amazon. When a user asks an AI to book a hotel, it is Booking that processes the transaction behind the scenes. AI doesn't eliminate the middleman. It becomes its customer.
That doesn't mean the risk is zero. If LLMs one day develop the ability to process payments in 50 currencies, manage a network of 4.4 million hotel partners, and resolve disputes in 40 languages, Booking will have a problem. That day hasn't come. And the cost per booking for customer service fell 10% in 2025 thanks to AI, a factual scorecard that, so far, tilts decidedly in Booking's favor.
The price of the fog
At $4,400, Booking trades at roughly 17 times estimated 2026 earnings, versus a historical average of about 27 times excluding Covid. The FCF yield sits around 7%. Add EPS growth of roughly 15% a year and a dividend of about 1%, and the implied return tops 16% annualized, and that assumes the multiple never re-rates. If the P/E were to revert to even 20 to 22 times over three years, the re-rating alone would tack on 8 to 10 points per year.
Two indicators will settle the debate. First, room nights. That is the ultimate arbiter. As long as that number keeps climbing, Booking is taking share and the AI thesis stays theoretical. Second, EBITDA margin. If it compresses, it means Booking is spending more to defend its turf. Three straight years of expansion say otherwise, but the market has its reasons to doubt, and the fog around AI won't lift in a single quarter.
What is certain is the price you pay for that fog. A global leader growing at 15%, with expanding margins, massive shareholder returns, and a payments infrastructure nobody can easily replicate, trading at the lowest multiple of the decade. The market is handing you a generous margin of safety, so long as you can live with the fog not clearing overnight.
