There is no moat: Booking’s CEO on a $1 stock, agentic travel, and why AI still doesn’t move the numbers

Glenn Fogel joined Priceline in February 2000, after waiting to collect his 1999 Wall Street bonus. The Nasdaq peaked about a week later. The company had been worth some $30 billion shortly after its IPO; by the time he arrived it was around $15 billion, and within nine months the market cap was a couple of hundred million, the stock a dollar, and a reverse split the only way to stay listed. He did not leave. This is his 27th year. Split-adjusted, that $6 share came close to $6,000 last summer — more than a thousand times in a little over a quarter-century. Booking Holdings, the parent of Booking.com, Priceline, Agoda and the rest, peaked near a $180 billion market cap and still sits well into the hundreds of billions. Elad Gil frames the guest as the person who walked into a few-hundred-million-dollar Priceline and helped turn it into a $100 billion-plus travel group now being rebuilt around AI.

The line Fogel wants his 25,000 employees to hear is the one he also gave the *Financial Times*: there is no such thing as a moat. There is no shelter from innovation. The company has real advantages today. They can be gone tomorrow. The only way to win for long is to keep inventing new services and to fight for the customer every day. From anyone else the sentence can sound like a pose. From someone who watched $30 billion become a few hundred million, it is a working rule.

The path to that seat was not a straight Wharton-to-CEO line. He graduated Wharton in finance, then started in Morgan Stanley’s MIS group, hanging tapes on IBM 3084s in a data center, then operating the mainframe, then writing code — and decided the work was not a career. Friends from undergrad were becoming bankers, so he went back for a Wharton MBA and a Harvard Law JD and landed on Wall Street. In 1995 the bank was acquired and most of the bankers were fired, including him. He is careful to say not everybody: some people were chosen to stay. He was not. The lesson he has carried ever since is what it feels like to sit on the other side of that table, and therefore how to do a layoff right and how to do it wrong.

Unemployment arrived in a pile. His father had died not long before, then his grandmother, then the dog. He was in his early thirties, single, and he finally wrote the novel he had always wanted to write. No agent picked it up. A blind date with a woman who had been an editor at Random House did not get the book published; it did get him a wife and two children, which he calls a wonderful life. She eventually said the relationship would not go forward unless he got a job. A law-school friend at Morgan Stanley put him in as head trader for Barton Biggs, a Wall Street legend. He had never traded anything. He did it for years and did not like it.

The late-1990s internet boom was the exit. The only real internet company of consequence on the East Coast, in his telling, was Priceline, and it had a corporate-development seat that fit a former banker who had also done IT. He took the offer and still insisted on waiting for the 1999 bonus, paid at the end of February 2000. Then the Nasdaq peaked. He jokes that going long the internet a week before the top proved he should not have been a trader. The unpublished novel is a closed chapter. The marriage and the 27-year stay are not.

Gil asks what the current AI wave should learn from that internet era. The numbers then were smaller; the optimism and the backlash rhyme. OpenAI and Anthropic are rumored, he notes, at $30 to $50 billion revenue run-rate each, while other firms can be valued at $10 billion with little or no revenue. In 1999 about 450 companies went public; another 450 went in the first months of 2000; perhaps 500 before that — on the order of 1,500 IPOs, of which at most two dozen survive. Those were the companies perceived as the strongest. Will AI repeat the ratio?

Fogel will not guess. Speculative booms, he says, have been a feature of the last 150 years: the California Gold Rush, the Detroit auto boom, people and money rushing in, a lot of innovation, a lot of lost capital, and still some firms of real economic value. That pattern is not new, and this time will not be an exception. On whether a founder should keep going or sell, he also refuses a general rule. Early Priceline, he admits, would have been happy if somebody had made an offer. The questions that matter are local: how confident are management and the capital, and what are you actually trying to do — build something that matters, or make money? The second is not a sin. It does have to be named. An American male’s life expectancy at birth is about 78, and it rises the longer you live. How do you want to spend the years?

When next-generation models dropped, the market decided agentic commerce would route around online travel agencies, and Booking’s stock took a hit. OpenAI had put checkout inside ChatGPT, with travel among the main use cases, then killed the feature; Booking rose about 8 percent on the news. Later, when OpenAI said it did not plan to be merchant of record and would not keep pushing in-app commerce that way, the worry faded further. Fogel does not want the cancellation over-read. People who do not know an industry, he says, look at travel and conclude AI will do it and the incumbents will be worth nothing. He wants the camera pulled back.

A business exists to do something of value for customers. Booking sits in a marketplace with two of them: travelers and partners. Large language models are, in his view, an extraordinary tool for making that job easier, cheaper, and better — not a reason the middle disappears. Loading inventory into a database is something anyone can do, and it is worth almost nothing. What Booking actually runs is thousands of people in conversation with hotels and property managers: how do we make your business better, where do you need demand, where are you hurting, what should change in your systems. “It’s so much more complex than, I believe, many other people who look at this industry from afar.”

The third layer is regulation. Travel is not banking, not airlines, not drug discovery, and it is still highly regulated, more so outside the United States, with the stack thickening rather than thinning. To be merchant of record you have to live inside those rules. Scale is what makes the compliance bill payable. His advice to anyone who thinks they can walk in and knock the giants over is to understand the business before committing capital.

Complicated family travel, Fogel says, is painful enough that people start planning and quit. The wealthy hire human concierges. Everyone else dumps the work on a spouse. He admits he and his wife still argue about whose turn it is. The pitch for agents is that they become that concierge for people who cannot staff one: a system that knows you, never forgets, searches a near-infinite set of permutations, goes down a dead end and backs up. People still want some agency. On a New York–Chicago business hop, a human assistant — or an agent — can just book it. On a messy trip, most people at least want to confirm.

A few nights before the taping he put a real itinerary into Penny, Priceline’s agentic assistant. He and his wife wanted the front of the plane; the young adults he is still paying for could sit in the back; one person had to originate in a different city; they would land in Europe in a city that was not the real destination, so the question was hotel that night or push on; restaurants; and a pile of frequent-flyer miles versus cash, by airline and by segment. Penny asked how many miles he had on each carrier and walked the flights. The plan matched what he had already suspected: miles for the two of them up front, cash for the kids, a hotel in the arrival city, a show the next day. He calls the back-and-forth beautiful.

The sharper point is failure. Weather and mechanicals cancel flights with nobody to blame, and travel is dominoes: one tile takes the rest. Customers need a single point of contact who can fix the chain. His goal is a system that predicts the break and rewrites the plan before it happens. Gil’s team had heard that Penny’s adoption doubled every month for several months, with a lift in conversion, a shorter path to booking, lower cancellations, and higher customer success. Fogel immediately pours water on it. Last year Booking did $186 billion of travel and a billion room nights. Penny is still really, really small, which is one reason they are not pushing it fully. He is living in another spreadsheet: tokens consumed, cost of that trip, ROI, whether the guest comes back, loyalty, lifetime value, and which model to use for which job while token prices keep falling.

Customer service is further along. AI picks up so nobody waits in a peak-time queue, and the old humiliation — a human who has to park you again while they find someone who can actually help — should go away. Cost per contact is down about 10 percent and satisfaction is up. The catch is that some people still want a person, and “what’s best for the customer” is the test, not whether AI can do the whole job. Capital follows the same sobriety. The year is more like $700 million of investment, not only the $550 million of cost savings they said they would put back into AI and platform, and not all of it is AI. First question: does reinvestment earn a sufficient return? Then acquisitions. If neither, give the money to shareholders, who will use it better. Over about twelve years they have bought back roughly 40 percent of the shares; Q4 alone was $3.6 billion of buybacks and more than $300 million of dividends, and Q1 shareholder returns hit a record $4 billion.

Gil, speaking for a Silicon Valley that sometimes overrates what agents will vaporize, points at a number: Booking closed 2025 with about 8.6 million alternative-accommodation listings — homes, rooms, odd places. If an agent is going to book a stay, it still needs a network. Fogel thinks Americans in particular under-appreciate the comparison. Globally, Booking’s listing count is not that different from Airbnb’s; on alternative-accommodation room nights it is about three-quarters the size of Airbnb, and that is before the much larger hotel business sitting on top. Like for like, that alternative business has grown faster than Airbnb for five years.

Then he turns the knife back to the opening line. Connectivity to those properties is an advantage against a pure-play AI front end. It is not a moat. Advantages disappear. The long-term win is still the agentic assistant he wants to build, plus the unglamorous partner work on the other side of the marketplace. Anyone who thinks the business is “get the inventory loaded” has not been in the rooms with the hotels.

Asked what he still wants from a life that already looks like a win, Fogel says he is lucky enough to do what he likes, and he likes this. Booking is not curing cancer. Travel still adds something: the mission is to make it easier for everybody to experience the world, other cultures, other people. People without a choice have to take the job that feeds a family. People with a little room should choose wisely, because the time does not come back. His private fear is the midlife version of his law-school classmates: they went because it was the path, became lawyers because that is what you do, stayed because it paid and was comfortable, and then wondered what they had done.

On jobs, he does not need a thought experiment. When Booking.com was acquired around 2004–2005 it sold hotels in more than 40 languages, which meant human beings writing every description and staffing customer service in every tongue. Machine translation arrived and those jobs disappeared. He still asks where those people went. History says technology also creates work. The live question is speed: disappearance and creation may not run at the same rate. Graduates already ask whether there will be an analyst seat, because that work is going to models. He keeps a second picture in his head: a 50-something driving an 18-wheeler across the United States, decent living, a sense of responsibility, and then the route is automated. How do you retrain that person, and what happens to a society that has lived through dislocations like that before?

Booking’s answer is upskilling he says he talks about every day with his chiefs: make people AI-literate even if the company cannot keep them, because the tools make them more employable somewhere else. He treats that as an obligation. In the short run it costs money; in the long run it is good for the firm and the person. Government retraining over the last 50 years, in his view, has not worked out so well. What he wants is a more honest public conversation, because a society that rejects technology out of fear will pay for it. In China, he says, nobody is having the argument about whether AI should be done; they are doing it. Gil adds that consumer surveys flip with the question: ask whether people love ChatGPT, Claude, and Gemini and they will pay more and use it for school; ask whether AI will take the job and the mood collapses, sometimes on purpose. Fogel, who is for democracy, wishes people who want votes would be a little more honest about what they are actually saying.