The Hilton-Marriott-Hyatt AI Premium, Triangulated From One Week of Earnings

I read the Hilton and Hyatt prints back to back at 6:40 Friday morning, the way you read a put-call ratio: not the absolute numbers, but the spread. Hilton on the tape Monday morning at 8 ET, RevPAR +3.6%, adjusted EBITDA $901M, and a Q&A section in which the words “AI,” “Anthropic,” and “use case” came up enough times that the sell-side analyst from Wells had to circle back twice to keep the model lined up with the script. Hyatt Wednesday at 7:01 ET, RevPAR +5.4%, an EPS beat that ran six cents over consensus, and an AI section that was shorter, narrower, and — read the way you’d read a 10-Q footnote — much more specific. Marriott files Tuesday May 6 before the bell. The third leg of the triangle drops in five trading sessions.

The contrarian thesis I want to put down before that print lands is this: the AI premium showing up in lodging-sector forward multiples is rational, not a narrative spread, and you can triangulate the wedge from one week of earnings calls. Hilton trades rich because the moat narrative is the cleanest. Hyatt trades cheap because the AI story is a margin story and margin stories rarely earn multiple expansion. Marriott trades in the middle because the AI story is a fee-growth story, and fee-growth stories earn the most when they’re being announced — which is what next Tuesday’s call is, in effect, structured to do.

The forward-P/E ladder, and why the spread is the chart

The numbers first, because the math has to do the work. GuruFocus had Hilton’s forward P/E sitting at 35.41x as of late April, against Marriott at 27.71x as the consensus from late March, with Hyatt trading at a meaningful discount to both on a forward basis — sell-side currently models Hyatt’s forward multiple in the high-teens once you adjust for the asset-light shift that hasn’t fully shown up in the FY27 numbers yet. The straight read on that ladder is that Hilton commands a 28% multiple premium to Marriott on a forward basis, and Marriott commands a roughly 40% premium to Hyatt on the same axis.

The standard story sell-side tells about that spread is RevPAR mix and asset-light ratio. Hilton has the highest fee-to-EBITDA ratio of the three; Marriott has the largest absolute fee base; Hyatt has the most owned-real-estate drag. That story explains maybe 15 of the 28 points of premium between Hilton and Marriott. It does not explain the rest. The rest is the AI premium, and the rest is what the two transcripts this week force you to mark.

Run the back-of-the-envelope. If you assume Hilton’s underlying free-cash-flow growth rate is 9% and Marriott’s is 10% — Marriott has slightly more unit-growth runway — then on a pure DCF basis Marriott should trade richer than Hilton, not cheaper. The fact that Hilton trades 28% richer despite a slightly lower underlying growth rate is the moat premium. And the moat premium, when you read the Monday transcript closely, is being earned in AI airtime.

Hilton’s moat story: 41 use cases and a named LLM partner

The most underpriced sentence on Monday’s Hilton call was Chris Nassetta’s enumeration of “41 distinct AI use cases” currently in deployment or pilot across the property network, anchored by the Anthropic-powered Hilton AI Planner that went live March 17 and is now embedded inside ChatGPT’s app catalog as a first-party search experience. The number 41 is not a number Nassetta would have put on a Q1 call if he didn’t want it to do specific work in analyst models. It’s the kind of disclosure that exists to anchor a moat narrative: not “we are investing in AI,” which every CEO says and which has no analytical content, but “we have forty-one specific deployments, an exclusive LLM partnership with a frontier-model lab, and a search-distribution channel inside the highest-traffic consumer AI app on the market.”

Translate that into multiple-expansion math. If you take the 41-use-case disclosure as a credibility signal that the AI overlay is real and the Anthropic partnership as a distribution-moat signal — Hilton is the only major lodging brand currently sitting inside the ChatGPT app drawer — then the premium-to-Marriott spread becomes legible. The 13 points of multiple premium that the underlying FCF growth doesn’t explain are the price the market is paying for an AI moat that the other two brands either don’t have (Hyatt) or have not yet announced (Marriott).

The risk to that story is also clean: it works as long as the ChatGPT-app channel is exclusive, and it compresses the day a competitor brand shows up next to Hilton in the same search drawer. The Hilton Q1 operator-side read I filed Monday afternoon flagged the same dynamic on the loyalty-program economics side — first-mover network effects work right up until they don’t. The 41-use-case number is the contract Nassetta is signing with the multiple. He has to keep producing those use cases at a credible cadence or the spread compresses on its own.

Hyatt’s margin story: vertical AI tools as the wedge

The Hyatt call was the most useful comparison set, because Hyatt is the one major branded operator that has chosen not to chase the moat-narrative version of AI and has instead built the margin-story version. Joan Bottarini’s section on the Wednesday transcript ran for less than three minutes of call time and named one specific deployment — the AI Wedding Guide that we covered as it rolled out — as the marquee operational AI product, alongside a much shorter list of back-of-house workflow tools focused on housekeeping routing, F&B inventory, and revenue-management overlay.

That is, structurally, a margin story. Hyatt isn’t claiming 41 use cases. Hyatt is claiming a small portfolio of vertical AI tools each of which is justified by an operational ROI on a specific cost line. The Wedding Guide is the cleanest example: weddings are roughly 4-6% of Hyatt’s group-room-night base by my rough cut, they’re the highest-touch group segment by sales-cycle length, and an AI overlay that compresses planner time even modestly drops the sales-and-marketing line by a measurable amount per booked event. That’s a margin contribution, not a multiple contribution.

The market is pricing that correctly. Margin contributions earn the EBITDA growth they generate; they do not earn multiple expansion on top of that growth. Hyatt’s forward P/E sitting at the bottom of the three-name ladder isn’t a punishment for not having an AI story — it’s the rational consequence of having an AI story that translates into the income statement instead of into a moat narrative. The +5.4% RevPAR print and the six-cent EPS beat tell you the margin story is working. The flat-to-down multiple tells you the market is collecting the gains in earnings, not in P/E.

That’s not a bad place to be if you’re a Hyatt shareholder. It is a place where the upside has to come from compounding margin, not from re-rating. Two different bets.

Marriott’s fee-growth thesis, and what May 6 has to do

The third leg is the one that hasn’t printed yet. Marriott reports Tuesday May 6 before the bell, and the consensus build on the call is structured around three pieces of forward guidance that have been telegraphed in prior investor materials but not yet confirmed in a quarterly print: the rollout of natural-language search across Marriott.com and the Bonvoy app by end of Q2 2026, the integration of AI agentic tooling into the central reservation system, and the leveraging of the Bonvoy membership base — last disclosed at nearly 283 million members in prior Marriott investor materials — as the data substrate for personalization.

If the May 6 call confirms those three items at the cadence the buy-side has built into their models, then Marriott’s fee-growth thesis gets the second leg it needs to defend its 27.71x forward multiple. Bonvoy’s 283-million-member base is the single largest first-party loyalty dataset in lodging, and a credible AI personalization overlay on top of that dataset is, in dollar terms, the largest fee-growth opportunity any of the three brands have available to them. Marriott has been signaling that direction for several quarters now, and Tuesday’s call is the natural moment for it to land formally inside the quarterly framework rather than as a side-deck disclosure.

The risk on that thesis is execution timing. If the natural-language-search rollout slips out of Q2 — and rollouts of this scale across a 9,000-property network do slip — then the fee-growth story becomes a 2027 story rather than a 2026 story, and the 27.71x multiple has to defend itself on the existing base for another two quarters. That’s a real downside scenario, and it’s the one I’d be reading the Tuesday call for: not the headline RevPAR number, but the language around the Bonvoy rollout schedule.

The bet: the spread is rational, here’s what compresses it

Read against the case-against-the-AI-premium thesis we ran in March and the restaurant-tech parallel, the lodging-sector version of the AI premium passes the rationality test. The spread between Hilton and Marriott — 28 points of forward P/E — is roughly proportional to the difference between “moat narrative with named LLM partner and 41 disclosed use cases” and “fee-growth thesis pending confirmation on a forthcoming print.” The spread between Marriott and Hyatt — another 40-plus percent — is roughly proportional to the difference between “fee-growth on 283 million loyalty members” and “margin contribution on vertical operational tools.” Both spreads are defensible on first principles.

Here’s where I’d put the compression markers. One: the day a second major lodging brand announces an exclusive frontier-model partnership of the Hilton-Anthropic shape, the Hilton-to-Marriott spread compresses by half. The moat narrative requires exclusivity, and the moment exclusivity goes, the premium goes. Two: if Tuesday’s Marriott call ships clean confirmation of the Q2 natural-language-search rollout and quantifies even a directional revenue contribution from the Bonvoy AI overlay, the Marriott-to-Hyatt spread widens rather than compresses, because the fee-growth story upgrades from “pending” to “live.” Three: if Hyatt’s next print delivers another margin beat without a moat upgrade, the bottom of the ladder stays where it is. The market does not pay for execution it can already see in the EBITDA line.

The contrarian risk to this whole framework is the obvious one. If AI in lodging turns out to be a commodity overlay — the way property-management software became a commodity in the 2010s — then all three spreads compress toward zero and the entire premium evaporates. That’s the bear case, and it’s the one the against-the-AI-premium piece walked through carefully. I’d take the other side of it for one specific reason: Hilton’s 41-use-case disclosure is the kind of number a CEO puts on a call only if he intends to be measured against it. The fact that Nassetta was willing to anchor the multiple to a specific operational count, rather than to a generic AI investment line, is the strongest signal in this week’s tape that the moat is real enough to be quantified.

The two prints are in. The third lands Tuesday. The wedge is the chart.

— Marcus edits The Bottom Line for TableTransfers. Tips: ma@tabletransfers.com.