Inside Marriott's Bonvoy AI overhaul: how 271M loyalty members became a co-branded card royalty machine

The Marriott Q4 2025 earnings call started at 8:30 a.m. Eastern on Tuesday, February 10. I had it on a second monitor with the transcript service open in a tab and the IR press release printed out next to my coffee. I was listening for one thing — what Anthony Capuano would say about food and beverage. He said almost nothing. The F&B line, if you read every page of the release, is mostly inside “other ancillary revenue,” and the operating commentary on the call moved past it inside ninety seconds. That is, on its own, a story.

The story that wound up dominating my notebook by the end of the call was a different one. The biggest number in the 2026 guide was not the EBITDA range. It was a single line, delivered in the prepared remarks and amplified by analyst questions later, about co-branded credit card fees. The number was 35%. As in: Marriott expects co-branded credit card fees to grow 35% in 2026.

That is not a hospitality KPI. That is a payment-network KPI. It is the kind of growth rate you see in the financials of a co-brand issuer in the second year of a re-priced contract, or in the bookings line of a marketplace business that just turned on demand-side advertising. It is not what fee revenue from a hotel-brand royalty stream usually looks like.

Read alongside what Capuano said about Bonvoy crossing 271 million members and what he said about the AI overhaul moving through Marriott’s three core platforms — property management, central reservations, loyalty — the framing of the company changes. Marriott is not a hotel chain that runs a loyalty programme and licenses some credit cards on the side. Marriott is a 271-million-member data business that uses a real-estate brand portfolio and a co-branded card programme as its two principal distribution and monetisation engines. The AI overhaul is the infrastructure that lets the data business actually run.

A note on methodology before I start, because the rest of this piece is a case study and case studies live or die on what’s verifiable. Everything here comes from public-record sources: the Marriott Q4 2025 earnings release on the company’s IR page, the Q4 2025 earnings call transcript published by The Motley Fool on the morning of February 10, the SEC EDGAR filing of the same release, and analyst commentary on the print. There is no interview with anyone at Marriott. Where I quote Capuano or CFO Leeny Oberg, the quote is from the transcript and I have flagged the source. Where I’m reading between the lines — about positioning, about what an operator should expect from this stack — I say so.

This is the long case study off the Q4 print. Five sections: the headline that wasn’t, the Bonvoy scale economics, the three-platform AI overhaul, the asset-light reframe, and what an operator partnering with Marriott should expect through 2026. I’ll close with a falsifiable prediction tied to the next Q1 print.

The headline that wasn’t

The headline that the wires ran with on February 10 was the EBITDA print. Fourth-quarter adjusted EBITDA of $1.4 billion, up 9% year-over-year. Fourth-quarter fee revenues of $1.4 billion, up 7%. Full-year 2025 adjusted EBITDA of $5.38 billion, up 8%. Full-year adjusted EPS of $10.02, up 7%. All of these were inside the consensus range. None of them moved the stock more than a percent in either direction in the first hour. The market had priced this in.

The headline that should have run is the 2026 guide. Marriott guided 2026 adjusted EBITDA growth of 8% to 10%, capital returns to shareholders of more than $4.3 billion, and — the line that did the work — co-branded credit card fee growth of 35% for 2026. That last number, on a $1 billion-plus base of co-branded card fees, is roughly $350 million of incremental fee revenue from one line item. To put that against the rest of the P&L: Marriott’s total fee revenue for the full year 2025 was about $5.7 billion. A single line in the fee stack is going to contribute the equivalent of about six percentage points of total fee growth in 2026, on its own, before RevPAR moves a basis point.

The mechanism is a re-priced co-brand contract. The tikr analyst note on the print called it out directly: Marriott amended what had been “a long-standing contractual limitation on its royalty rate” with the two card-issuing partners, Chase and American Express. Oberg, on the call, was careful to put the legal frame around it: “The increase in the royalty rate is supported by GAAP-required valuation that were performed by third parties when the credit card deals were signed.” Translation: the new rate is supported by third-party fair-value work that the issuers signed off on. Translation of the translation: the issuers agreed, in writing, that 271 million Bonvoy members and the spend velocity inside that base are worth more per dollar of card spend than the prior contract priced them at.

That is the moment a loyalty programme stops being a marketing line and becomes a balance-sheet asset. The card-issuer royalty rate is the externally validated price of one Bonvoy member’s annual card spend. When that price goes up by enough that the line item grows 35% in a single year, the equity story changes. You can run a discounted-cash-flow on a hotel company and get a number. You can run a DCF on a payment-network royalty stream and get a higher multiple. The market spent the rest of the trading day February 10 figuring out where between those two multiples Marriott should sit. The stock had already been up 51% over the prior twelve months going into the print, per tikr’s analyst note, and the post-print question was whether that re-rating had run out of room. The 35% line is the argument that it hasn’t.

For operators reading this, here is the implication: the Marriott corporate priority for 2026 is not RevPAR. RevPAR is guided at 1.5% to 2.5% — call it flat in real terms. The corporate priority is the card-fee line. Everything that drives card-fee growth — Bonvoy enrolment, Bonvoy engagement, Bonvoy spend velocity, the AI infrastructure that personalises the offers that drive card spend — is going to get budget, attention, and executive air-cover in 2026. Everything that does not is on a longer leash.

What 271 million members actually means

Capuano put the number in plain English in his prepared remarks: “43 million new members joined Bonvoy, propelling the membership base to 271 million members worldwide at year-end.” That growth rate — 43 million net adds in a single year on a base of 228 million — is roughly 19% annual. There is no other consumer loyalty programme of comparable size growing at that rate. The closest comparison is the airline ecosystems (AAdvantage and MileagePlus are each somewhere north of 100 million; Delta SkyMiles is the same order of magnitude). Bonvoy is now larger than the active user bases of most consumer fintech apps that have IPO’d in the last five years.

Scale is the variable that determines whether AI personalisation in a loyalty programme works as a business. Below a certain threshold, the cost of building the personalisation infrastructure exceeds the incremental gross margin from the personalisation itself. Above the threshold, the infrastructure is amortised over enough revenue-generating member sessions that the unit economics flip. The threshold is not a published number but you can back-of-the-envelope it. If a personalised next-best-offer engine costs $50 million per year to operate and lifts the average member’s card spend by $5, you need 10 million active members for it to break even, and 50 million active members for it to be a real margin line. Marriott has 271 million members. The infrastructure cost is fixed; the spend lift is multiplicative.

What I want to flag here is the second-order effect that does not show up in the slide deck. When you operate a personalisation engine over a member base this large, the marginal cost of each additional model improvement falls. You can test a new recommendation strategy on 10 million members, ship the winning variant to 100 million, and quantify the spend lift inside a quarter. The model improvement cadence inside a 271-million-member loyalty platform is structurally faster than inside a 30-million-member programme. The competitive moat is the experimentation throughput.

That is the asset Marriott bought when it consolidated Starwood Preferred Guest into Bonvoy in 2018. It was not visible in the SPG-era equity story, when loyalty programmes were marketing expense lines. It is visible now, in a 2026 guide that names a single co-branded card royalty line growing 35%.

The Bonvoy data asset is also the input to the AI overhaul. The reason the three-platform replatforming matters is not the platforms themselves. Property management systems, central reservations systems, and loyalty platforms are not, in 2026, technically difficult engineering. They have been built and rebuilt at every major hotel chain for forty years. The reason Marriott’s replatforming matters is that the new architecture, by all public accounts, is being built to surface the loyalty data into both the booking flow and the on-property experience. If that works, Bonvoy stops being a programme that members opt into and starts being the operating system that the entire stack runs on. The card-fee line is the early financial signal that the architecture is starting to deliver on that promise.

The three-platform overhaul

Capuano gave the AI overhaul exactly one line of strategic framing on the call, and it is the line that needs to be quoted in full to read it right: “We see AI as an opportunity to potentially redefine the customer acquisition paradigm that has governed our industry for the past several decades.”

That is the most ambitious public statement any major hotel CEO has made about AI in the last two years. It is also, importantly, a statement about distribution. The “customer acquisition paradigm that has governed our industry” is the OTA-and-search-engine paradigm — the one where Booking.com and Expedia and Google sit between the hotel and the guest and extract a take rate that has crept up over fifteen years. The implicit claim in Capuano’s framing is that AI changes the geometry of that funnel and that Marriott — by virtue of having a 271-million-member loyalty programme and a card programme that sits on top of it — is positioned to capture the change rather than be disintermediated by it.

I covered the mechanics of the replatforming in detail in the Marriott AI case study — eight thousand five hundred properties, the franchise-versus-corporate-mandate question, the $440 million 2026 technology budget, the three concurrent platform migrations. The thing to add from the Q4 call is the order. Capuano confirmed the sequence: property management is the most mature project, the central reservations migration is mid-stream, and the loyalty platform is being rebuilt around the new data architecture.

The order matters. Property management is the system that captures stay-level data. Central reservations is the system that captures search and booking data. The loyalty platform is the system that connects the two, attaches them to a member identity, and exposes the unified record to the personalisation and offer engines. You cannot build the loyalty-platform-driven personalisation engine — the one that drives the card-fee growth — without first having clean PMS and CRS data flowing into the loyalty record. The replatforming order is the order in which the data plumbing has to ship.

That is also why the natural-language search rollout is not landing in 2026 as a finished consumer product. The first-half-2026 timeline that Marriott telegraphed for natural-language search on Marriott.com and the Bonvoy app is conditional on the central reservations migration completing enough that the search front end can hit the new inventory and pricing APIs. If the CRS migration slips, the natural-language search rollout slips. Watch the Q1 print for the timeline update.

There is a parallel piece of architecture I haven’t seen named directly in the public record but that has to exist for the card-fee guide to hit: a real-time offer engine that takes the unified Bonvoy profile and decides, at the moment of a stay or a booking or a property check-in, which co-branded card offer to surface and to whom. The 35% card-fee growth is a function of two variables — the new royalty rate on existing spend, and incremental new-card-account acquisition. The royalty-rate piece is mechanical. The incremental-acquisition piece requires that the personalisation layer actually move the offer-acceptance needle. That is the AI-meets-card-marketing intersection that the replatforming has to deliver into, and the next twelve months will tell us whether it does.

The asset-light reframe

The line on the call I want to spend the rest of this section on came from Leeny Oberg, the CFO. It is the cleanest statement of the equity story I have heard from a hotel CFO in three years: “We were pleased that with the power of our strong cash-generating asset-light business model and our disciplined investment approach, we returned over $4 billion to shareholders.”

Asset-light is not a new framing for hotel-brand companies. Marriott, Hilton and Hyatt have been telling the asset-light story for a decade — own no real estate, license the brand, collect a percentage of revenues, return the cash. What is new in the 2026 framing is the part that goes after “asset-light.” Read the prepared remarks and the Q&A together and the framing has shifted. The 2026 capital allocation is $4.3 billion to shareholders, the largest capital priority on the slide. The 2026 capex line, including the technology investment, is materially smaller. The cash generation comes from a fee-revenue base that is growing inside a 7% to 9% range, plus a co-branded card royalty line that is growing 35%. The reinvestment requirement to keep the business growing is the technology budget — a fraction of the cash being returned.

That is the financial profile of a software company, not a real-estate-adjacent operator. The five-year EBITDA-to-free-cash-flow conversion at Marriott is now consistently above 70%. The capital intensity required to add a marginal hotel property is essentially zero — the franchisee or the third-party owner puts up the building, Marriott licenses the brand and the technology stack and collects the fee. The capital intensity required to add a marginal Bonvoy member is the cost of the digital enrolment flow. The capital intensity required to add a marginal co-branded card account is paid by the issuer.

The pipeline number underscores the brand-licensing leverage. Marriott ended 2025 with a development pipeline of approximately 610,000 rooms. None of that pipeline is on Marriott’s balance sheet. It is on the balance sheets of the owners and developers who have signed franchise or management agreements to build under one of Marriott’s brands. The fee stream from that pipeline lands on Marriott’s P&L over the next five-to-seven years as the rooms open. The capital that built the rooms came from someone else.

Stack the three components together and the equity story is: a fee-revenue base growing at low-to-mid-single-digits driven by RevPAR and net unit growth; a co-branded card royalty stream growing at 35% in 2026 on a re-priced contract; a 271-million-member data asset that compounds annually at 19%; a $4.3 billion-plus annual capital return; and a reinvestment requirement that is the technology budget. That is not a hotel operator. That is a data-and-licensing business with a hotel brand portfolio. The OpenTable–Booking Holdings comparison I drew in the OpenTable column — where the parent-company strategy reveals what the operating brand is actually for — is the right lens here too. Marriott Bonvoy is the operating brand. The parent-company strategy is the data-licensing economics.

There is one piece of the asset-light story that is not in the 2026 guide but that I want to flag for the Q1 follow-up. The luxury demand commentary on the call was sharper than I have heard from a hotel CEO in a year. Capuano said, in the prepared remarks: “When you look internationally, there is an almost insatiable demand for luxury. We’re seeing that across many of our markets.” Inside a 271-million-member loyalty programme, the high-spend luxury cohort is the highest-margin sub-segment, and it is also the cohort most likely to carry the premium tier of the co-branded card. The two are joined at the hip in the unit economics. If the luxury demand commentary is right and Marriott can capture more of it at higher-tier rates, the card-fee guide for 2027 — if the 2026 number lands — has additional room.

What an operator partnering with Marriott should expect over 2026

If you operate a property under a Marriott flag — managed, franchised, or somewhere in between — here is the read of what 2026 looks like for you given the priorities the Q4 print set.

The corporate spend will keep flowing into the technology stack, and you will see the bill for it. The $1 billion-plus annual technology investment, roughly 40% of total capex, is shared across the brand. Properties pay into the system funds that finance the replatformings. Expect the technology fee line on your monthly billings to keep ticking up through 2026. The PMS migration is the most disruptive piece — properties on the older systems will be on a migration schedule, and the migration itself is operationally costly. Plan for it.

The card-fee growth target will shape how Bonvoy enrolment and engagement are pushed at the property level. Bonvoy enrolment at check-in is going to become a more prominent corporate KPI in 2026 because the card-fee growth requires the member base to keep growing. Expect more training, more enrolment incentives, more friction-removal at the front desk. If you can hit your enrolment targets, you’ll be in a better position with your area management. If you can’t, you’ll hear about it.

The natural-language search rollout will reshape how your property is discovered. When the first-half-2026 rollout ships, the discovery surface on Marriott.com and the Bonvoy app changes. Properties whose listings are thinly populated — short descriptions, generic photos, sparse amenity tags — will surface less often in the AI-driven results. Properties whose listings are rich, structured, and differentiated will surface more. Audit your listing content now. The corporate team is not going to do it for you.

The card-offer surfacing will land at the property level eventually. I don’t have a public timeline for this, but the architectural logic of the personalisation engine implies that, at some point in the next eighteen months, the co-branded card offer that surfaces to a guest at check-in is going to be a function of the personalisation engine’s read of that guest’s profile rather than a static fall-back offer. When that ships, property-level enrolment economics change. The corporate team will likely run pilots first; ask your area director which properties are on the pilot list.

Food and beverage gets the air it has been getting. I want to be honest about this because the brief asked me to be a case study on Marriott’s strategy, not a defence of F&B’s seat at the corporate table. F&B was not a meaningful part of the Q4 call. It is not a meaningful part of the 2026 strategic priorities as the corporate team has framed them. The earlier Marriott AI case study covered the F&B-specific deployments — Winnow at 53 European hotels, Preferabli at the Napa Valley Marriott, IRIS mobile dining at 1,500-plus properties, SevenRooms as the preferred restaurant-tech vendor — and the rate at which those programmes expand will continue to be a function of regional and property-level economics rather than corporate mandate. If you operate F&B at a Marriott property, the operator-side AI you get in 2026 will look more like a continuation of what you already had than a new wave of deployments.

The falsifiable prediction

The standard I hold myself to in a case study is that I have to commit to a number that the next quarter can prove or break. Here it is.

Marriott’s Q1 2026 earnings call will land on or about May 5, 2026. On that call, I expect the co-branded credit card fee line — usually broken out in the franchise-and-other-fee segment of the release — to show year-over-year growth of at least 28%, and more likely 32% to 38%. The 35% full-year guide implies that the first quarter of the contract re-pricing is in already; the year-over-year compare in Q1 2026 will include the first full quarter of the new royalty rate against a base that did not include it. The acceleration in the line item is the early proof of the thesis.

If the Q1 print shows card-fee growth below 25%, the 35% full-year guide is in trouble, and the equity re-rating I described above gets unwound. If it lands inside the 32% to 38% range, the data-business framing of Marriott is, in my view, the correct lens, and the next twelve months are about how fast the AI personalisation layer can compound the card-fee economics into adjacent revenue streams. If it lands above 38%, Marriott will have to revise the full-year guide upward at the Q1 call, and we should expect the stock to take another leg up in the days after.

That’s the prediction. I’ll grade it in May.

A last note on what this case study is and is not. It is a public-record read of one quarter and one set of strategic priorities from one hotel-brand parent company. It is not an interview with the CEO. It is not a deep dive into the AI engineering inside the new loyalty platform. It is not an operator manual for working inside a Bonvoy property. It is the read of what the financial print and the transcript and the IR release said, in combination, about where Marriott has decided its growth is going to come from in 2026 and what that decision implies for everyone working downstream of it. If the read is wrong, the Q1 print will say so. If it is right, the Q1 print will give us the first piece of evidence and the rest of the year will give us the rest. The point of writing case studies in February is that they get tested in May. See you then.

— Naomi covers hotel F&B for TableTransfers. Tips: naomi@tabletransfers.com.