Every Guided Metric Cleared, and the Stock Went Nowhere: Initiating Booking Holdings at Hold

Results vs. Consensus

Q3 2025 Scorecard

Only revenue and adjusted EPS carry a genuine published Street mark for this name; the operating metrics are guided by the company and tracked against that guide rather than against a consensus print. The second table is therefore the one management and the buy side actually score the quarter on.

Versus the Company's Own Guidance

Year-Over-Year Comparisons

Quarter-Over-Quarter Comparisons

Q3 is structurally Booking's largest quarter, so the sequential move is seasonal rather than analytical. It is shown because the mix shifts inside it are not seasonal: merchant gross bookings grew about four percentage points faster sequentially than total gross bookings, and agency shrank again.

Revenue

Revenue of $9,008 million grew 12.7% and cleared the high end of guidance by about four points, entirely on the back of the room-night beat. The composition is where it gets interesting. Merchant revenues grew 23.3% to $6,131 million while agency revenues declined 6.7% to $2,569 million, continuing a multi-year migration that has now taken merchant to roughly 68% of trailing four-quarter gross bookings from about 61% a year ago. This is not a demand story, it is a payments story: Booking is increasingly the merchant of record, taking the payment risk and the payment economics rather than passing the booking to the hotel and collecting a commission.

The offsetting drag is take rate. Revenue as a percentage of gross bookings came in at 18.1%, about 30 basis points below last year. Management attributed the decline to a heavier mix of flight bookings, which carry a far lower take rate than accommodations, and to increased merchandising contra-revenue, some of it tied to bookings made in prior quarters. Higher payments revenue partly offset both. The mechanical point worth holding onto: the Connected Trip strategy that management is most excited about is, in its current form, take-rate dilutive. Flights grew 32% and attractions grew close to 90%, and both dilute the blended rate even as they raise engagement.

Assessment: An 8% constant-currency quarter with a declining take rate is a perfectly respectable result for a business of this scale, but it is not the growth profile that justifies a premium multiple on its own. What has to carry the valuation is the margin and the share count, and this quarter both delivered.

Margins

Adjusted EBITDA margin of 47.0% was up 112 basis points year over year, and the sources of leverage are the ones an investor should want. Marketing expense, the most variable line in the model, grew 8.8% against 12.7% revenue growth, taking marketing to 4.7% of gross bookings from 5.0%. Management credited traffic-mix changes and lower brand marketing as a percentage of gross bookings, and noted that marketing and merchandising combined also levered. Adjusted fixed operating expenses grew 9.7%, which the CFO characterised as mid-single digits after normalising for currency, against constant-currency revenue growth of about 8%. Fixed costs growing slower than constant-currency revenue is the definition of operating leverage, and it is happening while the Transformation Program is still ramping.

GAAP tells a less flattering story, and deliberately so. Net income margin fell 100 basis points to 30.5% because GAAP absorbed the $457 million KAYAK impairment and $105 million of transformation costs. The effective tax rate was 20.8% against 12.3% a year ago, though the prior-year figure was flattered by a $250 million Tax Act reversal and is not a clean comparison.

Assessment: The margin expansion is high quality and should persist. The Transformation Program has enabled about $450 million of annual run-rate savings against a raised $500 to $550 million target, and management is explicitly reinvesting roughly $170 million above baseline in 2025, which means the reported margin understates the underlying cost takeout.

EPS and Cash Flow

Adjusted EPS of $99.50 grew 18.6%, comfortably ahead of the $95.85 to $95.97 Street range. The bridge from adjusted EBITDA growth of 15.4% to adjusted EPS growth of 18.6% is almost entirely the 3.9% reduction in the diluted share count. GAAP EPS grew 13.5% to $84.41, the gap to adjusted reflecting the impairment, transformation costs and intangible amortisation.

Free cash flow is the one line that looks alarming and mostly is not. Q3 free cash flow of $1,371 million fell 40.3%, but this is a working-capital timing effect rather than an earnings problem. Booking runs a structurally negative working capital model in which deferred merchant bookings (customer cash held before the stay) swing violently with the booking calendar. The CFO attributed the sequential cash decline to a $2.4 billion reduction in deferred merchant bookings and other current liabilities as the summer travel season was consumed. On a nine-month basis, operating cash flow grew 4% and free cash flow grew 6%, and the release says so explicitly in a footnote.

Assessment: The nine-month figures are the honest read on cash generation, and they show cash flow growing far slower than adjusted EBITDA (6% versus 20% on a nine-month basis). That divergence deserves monitoring. It is explained today by the deferred-bookings unwind and by $1.5 billion paid to redeem high-coupon 2030 debt, but a persistent gap between EBITDA growth and cash growth is the sort of thing that is obvious only in hindsight.

Segment Performance

Booking does not report profit by operating segment, so the useful decomposition is by revenue type, by travel vertical, and by geography. All three moved in the same direction this quarter: toward merchant, toward non-accommodation verticals, and toward the United States.

Revenue by Type

Merchant

Merchant gross bookings grew 26% year over year and surpassed $123 billion on a trailing four-quarter basis, representing about 68% of total gross bookings against roughly 61% a year ago. The shift is deliberate and strategic rather than incidental. Being the merchant of record is what makes the Connected Trip technically possible, because bundling a flight, a hotel and an attraction into one transaction requires Booking to control the payment. It also generates incremental revenue and contribution margin dollars in its own right.

"Our merchant payments business is foundational to the Connected Trip, offers more flexibility for our travelers and partners and generate incremental revenue and contribution margin dollars for our business." — Ewout Steenbergen, CFO

The cost side is honest about the trade: sales and other expenses as a percentage of gross bookings rose slightly year over year, driven by higher payments expenses from the increasing merchant mix, partially offset by customer-service efficiencies.

Assessment: The merchant migration is the single most under-appreciated structural change in this business. It converts Booking from a commission broker into a payments company with a travel front end, which is both a higher-margin-dollar model and a meaningful competitive moat against an AI agent that can compare prices but cannot settle a cross-border payment, handle a chargeback, or absorb regulatory liability.

Agency

Agency revenues fell 6.7% and agency gross bookings fell 7.6% to $13.9 billion, the continuation of a steady decline visible across the nine-quarter statistical series. This is substantially a mix migration rather than a demand loss, as bookings that would once have been passed to the property for collection are now transacted on Booking's own rails.

Assessment: The decline is neutral to positive so long as it is cannibalisation rather than share loss. Nothing in this quarter suggests otherwise, but the line is worth watching as a check on the merchant story: if merchant growth ever decelerates while agency keeps falling, the migration explanation stops working.

Advertising and Other (KAYAK, OpenTable)

The smallest revenue line grew 14.5% to $308 million and is where the quarter's one genuinely bad piece of news sits. Booking recorded a $457 million impairment against KAYAK's goodwill and certain intangible assets, split between $180 million of goodwill and $277 million of intangibles covering trade names and supply and distribution agreements. The stated driver is unambiguous.

"The impairments were primarily driven by a reduction in the forecasted cash flows for KAYAK, reflecting its meta-search business being impacted by expected increases in customer acquisition costs." — Q3 2025 earnings release

OpenTable, the other asset in this line, had a better quarter: its loyalty program was relaunched as OpenTable Regulars with points redeemable on experiences and a new tier carrying benefits such as earlier notification of last-minute tables.

Assessment: Management framed the impairment as a KAYAK-specific problem and moved on, and no analyst pressed on it. That is a mistake on both sides. Meta-search is the part of the portfolio most directly exposed to AI-mediated discovery, and the company has now formally written down its forecast for that business because acquiring a customer is getting more expensive. That is the disintermediation thesis showing up as a number rather than as a question.

Travel Verticals and Key KPIs

Assessment: Every engagement metric moved the right way, and the direct-mix figures are the ones that matter most for the bear case. A business where the mid-60s of consumer bookings arrive without paying an intermediary for the click is far more defensible against AI-mediated search than one dependent on paid acquisition. The counterweight is that flight growth decelerated from the mid-40s to 32%, and rental cars grew just 2.1%, so the non-accommodation verticals are not uniformly accelerating.

Geography

All four major regions exceeded internal expectations. Europe and the United States grew room nights high single digits; Asia and Rest of World each delivered low double-digit growth. The United States is the story: growth accelerated to high single digits, the third consecutive quarter of acceleration, on both business-to-consumer and business-to-business strength. Management specifically flagged a step-up in the U.S. direct channel and said it believes Booking again outgrew the broader U.S. accommodations industry "in a meaningful way." The booking window in the U.S. normalised during the quarter, an improvement from Q2, though ADRs remained slightly lower and length of stay shorter than a year ago.

Global diversification did visible work: strength in corridors such as Canada to Mexico and Europe to Asia offset softer inbound demand to the United States.

Assessment: The U.S. acceleration is the most valuable thing in the quarter that is not in the financials. The U.S. is Booking's weakest large market by brand awareness and its most competitive, and three consecutive quarters of acceleration alongside rising direct mix is evidence that the brand investment is compounding rather than renting share.

Key Topics & Management Commentary

Overall Management Tone: Confident and unusually relaxed, with the CEO leaning into the competitive and AI questions rather than deflecting them. The posture on generative AI was notably more assertive than defensive: management repeatedly framed AI as a net competitive advantage given Booking's scale and proprietary data, and dismissed the hotel-direct disintermediation scenario as "an overblown threat at this time." The one place the commentary went thin was the KAYAK impairment, which was disclosed in the release, unmentioned in the prepared remarks, and never raised in Q&A.

1. The Room-Night Beat and the Booking-Window Pull-Forward

Room nights of 323 million grew 8.2% and beat the high end of guidance by nearly three percentage points. The CFO was explicit that a meaningful portion of the outperformance was borrowed rather than earned: the booking window expanded beyond expectations, pulling demand forward into Q3 that would otherwise have landed later.

"This outperformance was helped by an expansion of the booking window beyond our prior expectation and what we experienced in the second quarter, resulting in more room nights being pulled forward into the third quarter." — Ewout Steenbergen, CFO

The Q4 guide of 4% to 6% room-night growth reflects the payback directly, with management saying they expect the booking window to be "less expended" in Q4.

Assessment: Credit to management for flagging this rather than letting the beat stand unqualified. But it does mean the correct way to read the quarter is on a blended H2 basis, where 8.2% and a 4% to 6% guide average to something closer to 6.5%, which is below the 8% long-term room-night trajectory the multiple assumes.

2. U.S. Acceleration and the Brand-Awareness Gap

The U.S. accelerated for a third consecutive quarter across both consumer and business-to-business channels. What makes the commentary credible is that the CEO declined to claim a catalyst, describing it instead as accumulated execution.

"There's nothing that's miracle. There's no magic bullet happening, et cetera. It's bringing the brand together, it's doing a product better." — Glenn Fogel, CEO

He was equally candid that the job is unfinished, saying brand awareness in the U.S. "is still not where I want it to be," particularly in alternative accommodations.

Assessment: An admitted brand-awareness deficit in the largest travel market on earth is a growth option rather than a flaw, provided the cost of closing it stays inside the marketing leverage the company is currently generating. So far it has.

3. The $457 Million KAYAK Impairment

Booking wrote down $457 million of KAYAK goodwill and intangibles, comprising $180 million against the reporting unit's goodwill and $277 million against trade names and supply and distribution agreements. The disclosed cause is a reduction in forecast cash flows driven by expected increases in customer acquisition costs in meta-search. The charge appears on its own line in operating expenses and is excluded from adjusted results.

Neither the CEO nor the CFO mentioned the impairment in prepared remarks, and no analyst asked about it across seven questions.

Assessment: This is the quarter's most consequential disclosure and it passed without comment. KAYAK is the portfolio asset whose economics depend most directly on being found in search, and the company has now formally lowered its own forecast for that business on rising acquisition costs. Management spent the call arguing that AI-mediated discovery is not a threat while simultaneously taking a writedown whose stated cause is that discovery economics are deteriorating. Both things can be true, since Booking.com's direct mix genuinely insulates the core, but the juxtaposition deserved more scrutiny than it received.

4. Generative AI as a Distribution Channel

Booking was in the first wave of apps in OpenAI's ChatGPT App Store, having earlier been a launch partner for the operator platform. Management framed this as optionality on where discovery begins rather than a change to where transactions settle, and named relationships with OpenAI, Google, Amazon and Salesforce.

"The number of leads that we're receiving from large language models relatively small, but it is growing. And probably over time, these 2 worlds might become more hybrid because we are seeing, of course, more AI being built into browsers at this point in time." — Ewout Steenbergen, CFO

The CFO also offered the most useful single data point on the topic: traffic from traditional search is still growing year over year.

Assessment: Small but growing, with traditional search clicks still rising, is the honest current state and it is neither bullish nor bearish. The important disclosure is what is missing: no absolute number, no percentage, and no economics. Until one of those exists, the AI channel question cannot be underwritten in either direction.

5. The Disintermediation Question and Why Management Dismisses It

Asked directly about hotel partners bypassing OTAs by partnering with generative search players, the CEO gave the most forceful answer of the call, anchoring on the observation that the same fear about Google never materialised.

"But this idea that, that is going to cause this giant shift, I just think that that's not the way the world is going to work. And again, proof is that it hasn't happened in the old day of Google… But I think that is an overblown threat at this time." — Glenn Fogel, CEO

The supporting argument is structural rather than rhetorical: the mid-60s direct mix, the Genius program covering more than 30% of the active base and a mid-50s share of room nights, and the operational complexity of payments, regulation and fulfilment that a language model does not solve.

Assessment: The Google analogy is genuinely strong and is the best argument the bulls have. Its weakness is that Google monetised by selling clicks back to the OTAs, which preserved the intermediary layer, whereas an AI agent that completes a transaction removes it. The direct-mix and loyalty defence is the real answer, and it is the metric to track quarterly.

6. AI Inside the Cost Base

The least-discussed and most immediately quantifiable AI story is on the expense line. The CFO disclosed that customer service costs fell in absolute terms year over year despite volume growth of close to 10%, alongside reduced contact rates, faster handling times and higher satisfaction scores.

"I think it's actually remarkable. If you look at our results this quarter, our customer service costs are down year-over-year, are down year-over-year in absolute terms despite volume growth of close to 10%. So the average cost per booking is coming down very rapidly and the satisfaction scores are going up." — Ewout Steenbergen, CFO

Partner-facing tools followed the same pattern, with Smart Messenger and Auto-Reply generating partner satisfaction gains against prior messaging tools. Management also reported lower cancellation rates and improved conversion from better search and discovery.

Assessment: This is where AI is already worth money to Booking, and it is a large part of why adjusted fixed operating expenses grew mid-single digits in constant currency against 8% constant-currency revenue. The market is spending its attention on the speculative revenue risk while the certain cost benefit accrues quietly.

7. Connected Trip: Growing Fast, Diluting Take Rate

Connected Trip transactions, defined as a trip including more than one travel vertical, grew in the mid-20s year over year and now represent a low double-digit percentage of Booking.com's total transactions. Supporting verticals grew faster still: flights up 32%, attractions up close to 90%, payment-related bookings up in the mid-20s. New flight supply partnerships with Ryanair in Europe and Southwest in the U.S. were added during the quarter.

The strategic payoff management claims is retention: travellers who book multiple verticals return more often and book direct more frequently.

Assessment: The engagement logic is sound and the growth rates are real, but the honest accounting is that Connected Trip is currently a take-rate headwind, since flights were one of the two named causes of the 30 basis point decline in revenue as a percentage of gross bookings. This is a deliberate trade of near-term rate for long-term frequency and direct mix. It is the right trade, and it should be modelled as such rather than assumed to be rate-neutral.

8. Genius Loyalty and the Direct-Channel Flywheel

Travellers in Genius Levels 2 and 3 made up over 30% of the active base and accounted for a mid-50s percentage of room nights over the trailing four quarters, both up year over year. More than 850,000 partners now participate. Management emphasised that the program is designed to be incremental for partners rather than demand-shifting, citing higher review submission rates, better property visibility and improved off-peak occupancy.

The channel metrics moved together: trailing four-quarter B2C direct mix in the mid-60s against the low 60s a year ago, and mobile app mix in the mid-50s against the low 50s.

Assessment: This is the core of the bull case and the direct answer to the AI bear case. Every point of direct mix removes a point of dependency on third-party discovery, whether that third party is Google or a language model. The cost of the program runs through merchandising contra-revenue and is never sized, which is the one thing preventing a clean assessment of its return.

9. Transformation Program Raised Again

The expected ultimate annual run-rate savings target was raised to $500 to $550 million from $400 to $450 million. In-quarter savings were approximately $70 million, in-year 2025 savings are now expected to exceed $225 million, and roughly $450 million of annual run-rate savings has already been enabled. Management expects aggregate transformation costs to approximate one times the run-rate savings, and incurred $105 million of such costs this quarter.

Roughly $170 million above baseline is being reinvested in 2025, funded by these savings plus ongoing operational efficiencies.

Assessment: A program tracking meaningfully ahead of its original target with a clean one-times cost-to-savings ratio is a credible piece of the margin algorithm. The reinvestment disclosure matters because it means the reported margin expansion understates the gross cost takeout, and it gives management a lever to protect margins if revenue growth disappoints.

10. Capital Return and the Balance Sheet

Booking repurchased $0.7 billion of stock during the quarter with $23.9 billion remaining under authorisation, and declared a $9.60 per share dividend payable December 31, 2025. Ending cash and investments were $17.2 billion against $18.2 billion at the end of Q2, the decline driven by the $2.4 billion deferred-bookings reduction, capital returns of roughly $1.0 billion, and $1.5 billion paid to redeem high-coupon debt originally due 2030.

The balance sheet carries a $4,736 million stockholders' deficit, a function of $52.2 billion of cumulative treasury stock rather than distress; total liabilities of $33,488 million include $15,997 million of long-term debt against $16,509 million of cash.

Assessment: The buyback pace of $0.7 billion was modest this quarter relative to the $23.9 billion authorisation and to the roughly $4.3 billion repurchased over nine months. Retiring expensive 2030 paper is a sensible use of the same dollars. The share count is down 3.9% year over year and is contributing about a quarter of adjusted EPS growth, which is worth naming explicitly rather than treating as organic.

11. Competitive Intensity Over a Multi-Year Horizon

Pressed on whether competitive intensity rises over the next two to three years given an Asian competitor's European ambitions, an accommodation competitor pushing into hotels, and AI as a new channel, the CEO reframed the question as the industry's permanent condition rather than a new development, and named the specific competitor the question implied.

"I remember once upon time everybody talked about, what about TripAdvisor? Well, where is TripAdvisor now?… And I know you're alluding to Trip.com, Ctrip in China. Yes, a very good competitor. And sure, now we have AI as a competitive threat, but also in that one a little different because that gives us incredible benefits too." — Glenn Fogel, CEO

Assessment: Answering a competitive question by naming the competitor rather than deflecting is a confidence signal. The substantive claim, that scale and proprietary data make AI a net advantage, is plausible for the core but sits uncomfortably beside the KAYAK writedown, where scale did not protect the economics.

Guidance & Outlook

The full-year raise is unambiguous and broad: every line moved up, and the Adjusted EBITDA margin expansion target was lifted from about 125 basis points to about 180 basis points. Guidance assumes recent rates hold, with the euro at 1.17. Management also stated that on a constant-currency basis, the latest full-year expectations sit above the company's long-term ambition of at least 8% gross bookings and revenue growth and 15% adjusted EPS growth.

Implied Q4 ramp: Nine-month revenue of $20,568 million against a full-year guide of about 12% growth on a roughly $23.8 billion 2024 base implies fourth-quarter revenue near $6.1 billion, or about 11% growth against Q4 2024's $5.5 billion. That sits inside the 10% to 12% Q4 range, so the annual and quarterly guides are internally consistent rather than one implying a stretch in the other. The same check on Adjusted EBITDA holds: nine-month $7,742 million plus the $2.0 to $2.1 billion Q4 range gives $9.74 to $9.84 billion, which against the guided 17% to 18% growth back-solves to a 2024 base of roughly $8.3 billion and a Q4 growth range of about 8.7% to 14.1%, matching the stated 8% to 14%.

Street at: Consensus entering the print sat at $8.71 billion of revenue and $95.90 of adjusted EPS for the quarter, both of which were cleared. For Q4, the guide's 10% to 12% revenue growth is the number that matters, and the deceleration in room nights from 8.2% to 4% to 6% is the figure most likely to anchor the debate into the next print.

Guidance style: Conservative and consistently so. Management beat the high end of all four guided metrics this quarter, and the CFO's own framing of the misses-to-the-upside was precise to the percentage point, which suggests the guide is set to be cleared rather than to be aspirational. The relevant implication is that the Q4 room-night guide of 4% to 6% should be read as a floor set against a known pull-forward headwind, not as a forecast of underlying demand deterioration.

Analyst Q&A Highlights

Drivers Behind the Third Consecutive Quarter of U.S. Acceleration

The call opened on the U.S., with the questioner seeking the mechanics behind an acceleration visible in both business-to-business and consumer channels. Management declined to identify a discrete catalyst and instead attributed the result to accumulated product, supply, marketing and brand investment, with the CFO adding that the U.S. direct channel specifically had stepped up in the quarter.

Q: "I was hoping to dig in on your U.S. acceleration in Q3. Could you talk about your B2B initiatives in the U.S. that you mentioned and maybe globally? And then it sounds like B2C also accelerated in the U.S., so any additional color on what you saw as the key drivers there would be great."
— Kevin Kopelman, TD Cowen

A: "So what we are seeing is clearly a payoff of our brand awareness that is getting stronger in the U.S., more familiarity and therefore, more customers coming now direct to us in the U.S. So that is really something that has seen quite a step-up in the third quarter, and we see that as a really positive trend. And ultimately, that is all the result of all the investments we have been making over the last number of quarters and years."
— Ewout Steenbergen, CFO

Assessment: A direct-mix-led acceleration in the hardest major market is higher quality than a promotion-led one, because it does not need to be repurchased next quarter. The absence of a named catalyst is a feature rather than an evasion here.

Economics of the OpenAI App Integration

The first of three AI questions probed what happens to unit economics if bookings migrate from direct or Google traffic into a language-model interface. Management conceded it is too early for data and pivoted to the argument that discovery may move while fulfilment will not. The CFO supplied the only quantification offered on the call.

Q: "I know you were early in the test program with OpenAI as well, but can you talk about your thought process heading into the app integration and what you're seeing in the early days? And just how should we think about economic impact if bookings were to shift from direct traffic or from Google?"
— Douglas Anmuth, JPMorgan

A: "So first of all, what we are seeing in terms of traditional search that we still see volume growth. So travel clicks that are coming to us from traditional search are still going up year-over-year. That, of course, might change over time, but I think that is an important data point. The other is the number of leads that we're receiving from large language models relatively small, but it is growing."
— Ewout Steenbergen, CFO

Assessment: The question about economics went unanswered, and reasonably so, because the economics are not yet knowable. What the exchange did establish is that the substitution has not started in the traffic data. That is a real, checkable fact and it should be re-checked every quarter.

Whether Hotel Partners Can Bypass the Platform Through Generative Search

The sharpest challenge of the call put the disintermediation scenario directly. Management rejected the premise on historical grounds, arguing the same dynamic was predicted for Google and did not occur, and pointed to direct mix and loyalty as the structural defence.

Q: "There's obviously a lot of noise in the market around some of your hotel partners looking to partner directly with some of the generative search players in order to perhaps increasing the bypass platforms like yourselves. I guess how do you contextualize this particular risk? And what tools do you think you have at your disposal to maybe mitigate this kind of disruption?"
— Lee Horowitz, Deutsche Bank

A: "So obviously, I don't disagree. Some people are going to go to a large language model. You'll see a hotel to go directly there. Sure. But I think that is an overblown threat at this time."
— Glenn Fogel, CEO

Assessment: A confident, unhedged answer, and the historical analogy has genuine force. The qualifier "at this time" is doing quiet work, and it is the correct qualifier given that the company simultaneously wrote down KAYAK on deteriorating acquisition economics.

Social Media Marketing Scale and the Asia Growth Engine

A two-part question sought to size the social-media performance-marketing channel and to understand where within Asia the dual-brand strategy is working. Management sized the social spend for the first time and declined to break out which platforms perform best, explicitly to avoid informing competitors.

Q: "First is social media, I think you mentioned kind of leaning into social media marketing. I think you've been talking about this for a year and maybe to — could you spend a little bit more time on that? Has that now become a material, let's say, a double-digit percent of your performance marketing coming from there?"
— Mark Mahaney, Evercore ISI

A: "In terms of spend, you should think about that it is a couple of hundreds of millions, which is meaningful, of course, a total number, but if you look at the total marketing spend for us, it's, of course, still a smaller percentage of the overall spend."
— Ewout Steenbergen, CFO

Assessment: A couple of hundred million against roughly $2.3 billion of quarterly marketing spend confirms social is still a rounding error at the corporate level. The strategic value is channel diversification away from search, which matters more for the AI debate than the current spend implies.

Measured Product Impact of AI on Conversion and Cancellations

A question on how the wave of new AI-driven product surfaces is affecting funnel metrics produced the most concrete operating disclosure of the call, covering conversion, cancellation rates and the customer-service cost line.

Q: "On the product side, look, seeing tons of innovation with AI Trip Planner, Penny, hotel search, AI Mode, concierge. I think you talked about the homepage is unique now per user. Talk to us about the impacts this might be having on either cancellation rates, conversion rates, things along those lines."
— Ronald Josey, Citi

A: "First, conversion, definitely with all the new tools that we're having, we're seeing that people have an opportunity to find easier what they are looking for can be more targeted, have a faster path to ultimately get to a booking and that is helping conversion levels. Then also what we are seeing at the same time is that cancellation rates are a bit lower."
— Ewout Steenbergen, CFO

Assessment: The most valuable answer on the call, and the one that reframes the AI question from threat to margin driver. Lower cancellations improve the quality of the gross-bookings line, and falling absolute customer-service costs on 10% volume growth is a durable structural gain rather than a one-quarter benefit.

Whether the Gap Between Gross Bookings and Room-Night Growth Can Widen

A follow-up sought to understand the durability of the wedge between gross bookings growth and room-night growth, which is the mechanical bridge between volume and revenue in this model. Management confirmed the intent for that gap to persist, naming the three inputs that drive it.

Q: "Do you think the gap between bookings growth and room night growth can grow?"
— Justin Post, Bank of America

A: "Yes. There's, of course, always a few things that go in the mix, take FX aside, what our ADR is doing. That is, of course, one element there. It's the growth in other verticals that is going in there. But generally, we would say, yes, because if we take room nights plus ADRs plus other verticals, we would like to, of course, to see that total gross bookings will grow faster than room nights on average over time."
— Ewout Steenbergen, CFO

Assessment: A useful commitment for modelling purposes, because it makes gross bookings growth structurally faster than room nights. It is also a reminder that the wedge is partly ADR, which is macro rather than execution, and constant-currency ADRs grew only about 1% this quarter.

Whether Competitive Intensity Rises Over the Next Two to Three Years

The closing question stacked three competitive vectors into one and asked how strategy evolves in response. Management's answer was the most expansive of the call, arguing the industry has always been maximally competitive and that AI is a net advantage at Booking's scale.

Q: "From a competitive standpoint, do you foresee competitive intensity picking up over the next 2 to 3 years? It seems as though a major Asia player has aspirations in Europe while an accommodation competitor is making a big push into traditional hotels and adjacencies. Meanwhile, consumers will have a new channel via AI tools to make travel all that much more accessible."
— Trevor Young, Barclays

A: "But I have to say that this business, this industry, and I take to the regulators all the time in Europe, has always been one of the most competitive of any industry than I've ever experienced… I actually see us in a much better position than we were years ago, decades ago."
— Glenn Fogel, CEO

Assessment: Confident and historically grounded, but it is an argument from analogy rather than from data. The verifiable version of this claim is the direct-mix and Genius-mix series, which is why those metrics deserve more weight in the model than the narrative around them.

What They're NOT Saying

1. Any framing for 2026: No preliminary 2026 algorithm was offered, not even a directional comment, despite the full-year 2025 guide now being effectively locked. The long-term ambition of at least 8% gross bookings and revenue growth with 15% adjusted EPS growth was restated, but restating a long-term ambition is not guidance.
2. Any absolute measure of language-model referral volume: "Relatively small, but it is growing" is the entire quantification of the single most debated risk to the business. No percentage of traffic, no absolute leads, no conversion or economics comparison against organic or paid channels.
3. KAYAK's revenue and profit contribution: A $457 million impairment was taken against an asset whose standalone economics have never been disclosed. Investors were told the forecast came down without being told what it came down from.
4. The economics of the OpenAI relationship: No disclosure of whether Booking pays for placement in the ChatGPT App Store, shares revenue, retains the customer relationship, or receives attribution data. The distribution relationship was announced; its terms were not.
5. The cost of Genius: The loyalty program's tiered discounts and perks run through merchandising contra-revenue, which management named as one of two causes of the take-rate decline. The program's cost is never sized, so its return on investment cannot be independently assessed.
6. Where take rate settles: Revenue as a percentage of gross bookings fell about 30 basis points and management explained the mechanics, but offered no view on the trajectory. Given that the two named causes (flight mix and merchandising) are both strategic priorities, the decline should be expected to continue, and management did not say so.
7. Rental cars: The vertical grew 2.1% and declined sequentially, and went entirely unmentioned in prepared remarks and Q&A while flights and attractions were highlighted repeatedly.

Market Reaction

• Pre-print setup: BKNG closed at $5,120.50 on October 28 ahead of the after-market release. The stock entered the print up just 3.1% year to date against a 17.2% gain for the S&P 500, down 7.7% over the trailing 30 days, and up 17.8% over the trailing twelve months. The 52-week closing range was $4,164.25 to $5,816.00, placing the pre-print level about 12% below the high.
• Next-day session: Shares opened at $5,200.00, a 1.6% gap higher, traded as low as $4,928.25 and as high as $5,220.00, and closed at $5,075.75, down 0.9% on the day. The full peak-to-trough range was 5.6%.
• Volume: 11.5 million shares against a 6.2 million 30-day average, or 1.9 times normal. Heavy participation, no directional resolution.
• Benchmark: The S&P 500 was unchanged on the session, so the 0.9% decline was idiosyncratic rather than market-driven.

A company that clears the high end of all four guided metrics, raises the full year across every line, and lifts its cost-savings target does not usually open up 1.6% and close down 0.9%. The reaction is the most informative thing about this quarter, because it says the print was not the variable being priced.

Three things explain it. First, the quality composition was visible immediately: 12.7% reported revenue growth against 8% constant currency meant the beat was substantially currency, and the Q4 guide embeds a further 5 points of FX benefit that everyone knows will reverse. Second, management pre-announced the payback, guiding room nights to 4% to 6% after an 8.2% quarter that they explicitly attributed in part to a pulled-forward booking window. A beat that borrows from the next quarter is worth less than a beat that does not. Third, and least discussed, the release contained a $457 million impairment whose stated cause is rising customer acquisition costs, which is the precise mechanism the AI bears have been describing, appearing in the financials for the first time.

The intraday pattern supports this reading. The gap higher was the algorithmic response to the headline beat; the fade through the session, on nearly twice normal volume, was the considered response to the composition. The stock's year-to-date underperformance against the index by roughly 14 points going into the print tells you this de-rating has been underway for most of 2025, and one clean quarter did not arrest it.

Street Perspective

Debate: Does generative AI disintermediate the online travel agency or entrench it?

Bull view: The bull case holds that discovery moving to language models is irrelevant because Booking's value is in fulfilment, not in being found. The mid-60s direct mix, the Genius program covering a mid-50s share of room nights, and the payments, regulatory and customer-service infrastructure are things a model cannot replicate, and the same fear about Google proved unfounded for two decades.

Bear view: The bear camp argues the Google analogy fails at the decisive point, because Google monetised by selling traffic back to the intermediary while an AI agent that completes a booking removes the intermediary. The bears now point to the KAYAK impairment as first evidence, since Booking itself has cut forecasts on rising acquisition costs in the asset most exposed to search.

Our take: The bulls have the better argument about the core and the bears have the better evidence about the periphery, and both are currently correct. The distinction that matters is that Booking.com's economics rest on direct and loyalty traffic while KAYAK's rest on paid discovery, so the writedown is a genuine signal about meta-search without being a verdict on the core. This debate will not be settled by argument; it will be settled by the direct-mix series and by an eventual disclosure of language-model referral volume, and until then the risk is real, unquantifiable, and a legitimate reason to demand a discount.

Debate: Is 8% constant-currency revenue growth the floor or the ceiling?

Bull view: The optimists note that 8% constant currency was achieved with a low-single-digit ADR contribution and against a strong prior-year comparison, and that the growth algorithm has multiple untapped levers: U.S. brand awareness that management itself calls inadequate, Asia growing at high single digits industry-wide with Booking targeting faster, alternative accommodations at 36% of room nights and rising, and attractions compounding near 90%.

Bear view: The skeptics observe that 8% is the bottom of the company's own "at least 8%" long-term ambition, that room-night growth decelerates to 4% to 6% next quarter, that the take rate is falling because the fastest-growing verticals are the least profitable, and that ADRs at roughly 1% constant currency provide no pricing tailwind.

Our take: The bears have this one on the near-term math. Blending the 8.2% Q3 room-night result with a 4% to 6% Q4 guide gives roughly 6.5% for the second half, and the take rate is declining by design. The offsetting truth is that gross bookings should structurally outgrow room nights, which management confirmed, so revenue growth need not track volume growth down. The realistic base case is high single digits in constant currency, which supports a good business and a market multiple, not a premium one.

Debate: Does the margin and buyback algorithm justify paying a market multiple?

Bull view: The bull argument is that 47% Adjusted EBITDA margins, structurally negative working capital, a $23.9 billion repurchase authorisation against a share count already falling 3.9% a year, and a cost program running ahead of target combine into a mid-to-high-teens earnings algorithm that requires only high-single-digit revenue growth to work. At roughly 22.5 times forward earnings, that is a market multiple for a well-above-market compounder.

Bear view: The bear response is that the earnings algorithm is flattered on both ends. Roughly four points of this year's adjusted EPS growth is currency and about four points is share count, leaving underlying operating growth in the low teens, and the currency contribution is a coin flip that has recently been landing heads. Nine-month free cash flow grew 6% against 20% adjusted EBITDA growth, which is not the cash conversion the bull case assumes.

Our take: The bears are right that the headline algorithm is overstated and the bulls are right that the underlying one is still good. Low-teens underlying earnings growth with a genuine buyback and 47% margins deserves roughly a market multiple, which is approximately what it trades at. That symmetry is precisely why this is a Hold rather than either a buy or an avoid. The free cash flow divergence is the number we would most want to see close before paying up.

Model Update & Valuation Framework

Valuation. At the October 29 close of $5,075.75, BKNG trades at approximately 23.0 times trailing twelve-month adjusted EPS of $220.69, about 22.5 times our FY2025 estimate of roughly $226, and about 19.5 times our FY2026 estimate of roughly $260 built on the company's own long-term earnings algorithm.

Fair value framework. We anchor on FY2026 earnings power and apply a 20 to 22 times multiple, which is a modest discount to the current trailing multiple and reflects the unresolved AI overhang against genuinely high business quality. That produces a fair value range of roughly $5,200 to $5,720, with a midpoint near $5,460. Against the current $5,075.75 that implies about 2% to 13% of upside, with a midpoint around 7.6%, which is approximately a market return and is the arithmetic behind the Hold.

What would change the rating. Upgrading to Outperform requires either a cheaper entry, which we would define as roughly 18 times forward earnings or below, or hard evidence that closes the AI question: a disclosed language-model referral figure that is immaterial, continued direct-mix expansion into the high 60s, or a quarter in which constant-currency revenue growth re-accelerates above 10%. Downgrading to Underperform would require the direct-mix series to stall or reverse, a second impairment in the discovery-dependent assets, or evidence that the take-rate decline is accelerating beyond the mix explanation.

Thesis Scorecard Post-Earnings

This is our initiation, so the scorecard establishes the pillars we will carry and grade in subsequent quarters rather than scoring a pre-existing thesis.

Overall: Thesis established. The three bull pillars are each confirmed by hard metrics in this quarter, which is why this is not an Underperform. Two of the three bear points are also confirmed by hard metrics in this same quarter, which is why it is not an Outperform. The business is executing; the price already reflects it; and the largest risk is genuinely unquantifiable today.

Action: Hold. Own it for the compounding, do not pay up for it here. We would become buyers on a de-rating toward 18 times forward earnings, or on evidence that the direct-mix flywheel is outrunning the discovery risk faster than the KAYAK writedown suggests.