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

It is Friday morning, May 1, and I have three monitors loaded with the same comp sheet rebuilt three different ways. Hilton’s Q1 transcript is open on the left. Hyatt’s Q1 transcript is open in the middle. On the right I have a placeholder slot for Marriott, which prints next Wednesday, May 6. I have been telling colleagues for two weeks that this cluster — three lodging majors reporting inside seven trading days, all of them dropping AI artifacts at the same time — was going to be the most useful read-across of the year for anyone trying to defend or attack the “AI premium” in hospitality multiples.

I was right about the cluster. I was wrong about the conclusion I expected to reach.

My base case going in was that the AI talk was theater — the kind of analyst-day vocabulary that gets you a half-turn of multiple expansion until somebody asks what you actually shipped. Two of these three companies have already disabused me of that view. Hilton and Hyatt gave the Street something specific enough to model. The third, Marriott, has been telegraphing its own specifics through preview channels for a quarter. And when I look at where the forward multiples sit right now, the spread is not random. It is, against my prior, a rational AI premium.

Let me show my work.

Three companies, three AI stories

Here is the cluster as I am pricing it on May 1, in one table, before I start pulling it apart:

Company	Report	Headline metric	AI artifact	Forward P/E
Hilton (HLT)	April 28 (reported)	RevPAR +3.6%, Adj EBITDA $901M	AI Planner GA on hilton.com (March 17), 41 use cases in test, named partners Google + OpenAI + Anthropic	~35x (GuruFocus trailing week)
Hyatt (H)	April 30 (reported)	RevPAR +5.4%, gross fees $333M (+8.6%), 2026 net income guide raised to $255–$350M	AI Wedding Guide	cheaper still
Marriott (MAR)	May 6 (anticipated)	Consensus RevPAR +4.2%, Adj EBITDA $1.4B (+15% y/y)	1,000th hotel transitioned to the new ecosystem; natural-language search expected by end Q2	27.71x (GuruFocus, late March)

The forward multiple on Hilton has been the conversation piece in every lodging chat I am on this week. At roughly a 35-handle on forward earnings against a 27-handle for Marriott, you are paying roughly 28% more per dollar of next year’s earnings for Hilton than for Marriott, and noticeably more than for Hyatt. That is the AI premium I want to interrogate. Is it earned, or is it the kind of thing that compresses the first time RevPAR misses by 30 basis points?

Why Hilton’s 41 use cases is the right brag

Chris Nassetta did something on the April 28 call that I think the sell-side under-weighted in its same-day notes. He named his model partner. In his words: “we did with Anthropic and Claude.” That is not a sentence I have ever heard a lodging CEO speak before. Plenty of operators have talked about “leveraging generative AI”; very few have stood up and said which lab they are paying.

The reason this matters for valuation is downstream. A named partnership creates a model dependency the Street can underwrite. You can ask follow-up questions about latency, cost per query, refresh cadence, and switching cost. You cannot do that with a generic “we are exploring AI.” Skift’s coverage of the same earnings is a good summary of how broad Hilton’s vendor surface is: Google, OpenAI, and Anthropic all in the mix, each with a different role. Hotel Dive’s read lands at the same place: this is a portfolio bet, not a single-vendor bet.

The “41 use cases in test” number is the second artifact I keep coming back to. Per Hilton’s commentary and the Phocuswire summary, the AI Planner that went GA on hilton.com on March 17 is one of 41 live AI workstreams. That number reads to me less like a brag and more like a moat narrative. 9,100-plus properties and 1.3 million-plus rooms across the system means that whichever of those 41 use cases hits has a deployment surface most independent operators cannot match. Hilton is essentially saying: pick whichever of these wins; we run the rails.

The valuation read across is straightforward. If you believe even three of the 41 use cases drive measurable conversion lift or owner-side cost takeout, you can justify a half-turn to a full turn of multiple expansion. The market is currently giving Hilton something closer to two turns. That is the part I want to argue with — but only after I credit the artifact.

Where I disagree with the Street: the 35x handle prices in execution risk near zero. I think the right number is closer to 32x. The AI Planner is real, but conversion-lift attribution at scale on hilton.com will take two full quarters of A/B data to defend, and the market will get one disappointing read along the way.

Why Hyatt’s wedding tool is the cheapest AI dollar in lodging

Hyatt did not try to out-Hilton Hilton on AI ambition, and I think that was the right call. On the April 30 call, Mark Hoplamazian put up RevPAR +5.4%, gross fees of $333M (up 8.6% year over year), raised the 2026 net income guide to $255–$350M, and walked the Street through the AI Wedding Guide — a single, vertical, high-intent tool for one of the highest-margin segments in group business.

If you are an analyst trying to value AI optionality in lodging on a discounted cash-flow basis, the Wedding Guide is the cleanest case study you are going to see this year. Weddings are a known revenue pool. Conversion is observable. Average booking value is high. The deployment surface is narrow enough that the engineering cost is manageable. And the Hyatt earnings room telegraphs that the company is treating it as a wedge into broader group AI, not as a one-off marketing toy.

My base case on Hyatt: the AI dollar there is the cheapest in the cluster, in the sense that the company is spending the least to produce the most legibly attributable AI revenue. The market has not priced this. Hyatt trades cheaper than both Hilton and Marriott on forward earnings. If the Wedding Guide does even half of what the company is implying, the multiple closes some of the gap to Marriott in the back half of the year.

Where I disagree with the Street: Hyatt is being valued as a smaller, slower-growing lodging story. The AI artifact reframes it as a higher-margin, group-tilted operator. That is not in the consensus model.

What Marriott has to prove on Wednesday

This is the part of the piece where I have to be careful, because the Marriott print is six days out and I am not in the business of writing previews that read like post-mortems. Here is what I am watching for on May 6, and what I expect the company to disclose.

Consensus has Marriott RevPAR at +4.2% and Adj EBITDA at roughly $1.4B, up about 15% year over year. Marriott is also expected to disclose progress on the technology ecosystem rollout — the multi-year cutover that has been moving properties off the legacy stack. The number that has been telegraphed through analyst-day and preview channels is the 1,000th-hotel transition milestone, with natural-language search across booking inventory expected by end Q2 2026. We covered the deployment shape of that program in Marriott AI deployment.

If Marriott confirms the 1,000th hotel and the end-Q2 natural-language search timeline, the read across is this: Marriott’s AI artifact is ecosystem cutover, not product launch. That has different economics. Ecosystem cutovers compound — every additional property on the new stack lowers the marginal cost of the next AI feature and raises the incremental fee economics. It is less visible than a consumer-facing AI Planner. It is also harder to fake.

The bear case is the obvious one. Marriott has been telegraphing this cutover for several quarters. If May 6 brings the 1,000th-hotel number and not much else — no natural-language search demo, no fee-revenue uplift attributable to the new stack, no owner-side cost takeout color — the 27.71x multiple is the right number, not a cheap one. The premium versus the historical lodging average is being paid for the ecosystem optionality. If the optionality is not visible by the back half, the multiple compresses.

What I will be looking at first on the Marriott deck:

1. Bonvoy member count. Tony Capuano said on the last call that Bonvoy ended March with “nearly 283 million members.” The directional read on the May 6 number tells you how much of the AI demand-gen story is real versus aspirational.
2. Fee growth decomposition. Gross fees up versus management-fees-only versus franchise. If the franchise line is accelerating because the ecosystem cutover is letting more independent owners convert, that is the AI dollar Marriott has been promising for two years.
3. Capex on technology. Up or flat is fine. Down is a red flag.

What this cluster does to the AI-premium debate

The honest answer is that this week of earnings has made me more comfortable with the AI premium in lodging than I was a month ago, and less comfortable with it elsewhere.

I have argued in two prior pieces — the restaurant-tech AI premium thesis and the case against the AI premium — that operators get a multiple bump for AI talk that is almost never earned at the level the Street is pricing. I still think that is true in restaurant tech and true in most of mid-cap travel software. The voice-agent space, which I wrote about in the Voice Agent Maturity Curve, is the cleanest example of premium without artifact.

Lodging is different. Lodging has three things working in its favor on this question. First, the deployment surface is enormous and observable — 9,100 Hilton properties, similar order of magnitude at Marriott, every booking surface measurable. Second, the AI use cases line up with metrics the Street already models (RevPAR, conversion, group revenue, fee growth). Third, the cluster is publishing artifacts on the same week, which means cross-comp is possible in real time. You do not get that in restaurant tech or travel SaaS.

So when I look at the forward-multiple spread — Hilton at the 35-handle, Marriott at the 27-handle, Hyatt cheaper still — I read it as the market pricing three different bets:

• Hilton: moat narrative, broad partner surface, consumer-facing artifact, highest premium.
• Marriott: ecosystem cutover, fee compounder, lower visibility, mid premium.
• Hyatt: vertical tool, margin story, narrowest surface, smallest premium.

All three can be right, and the spread between them is rational. That is not a sentence I expected to write on Monday.

My base case after the cluster

Hilton at 35x is mildly overvalued. Fair value, on my numbers, is closer to 32x. I would not short it — execution on the AI Planner has been clean enough that I do not want to be on the other side of a beat — but I am not adding here.

Hyatt at the current handle is the most interesting risk-reward in the cluster. The AI dollar is cheap, the group-revenue tilt is underpriced, and the guide-up on net income gives the back half a margin cushion. This is the one I would add to.

Marriott on May 6 is the binary. Confirm the 1,000th hotel and the end-Q2 natural-language search timeline and the 27x is reasonable. Anything less and the multiple compresses to 24-25x and the premium versus the lodging average evaporates.

The broader read for anyone valuing AI premiums in 2026: artifact-based pricing works. The companies that ship specific, named, measurable AI deployments deserve a multiple. The ones that do not, do not. This week was the first time I have seen three majors give the Street the artifacts on the same calendar page. That is the unlock.

I will be back next Wednesday night with the Marriott read.

— Oliver writes The Bottom Line on M&A and valuations. Tips: tips@tabletransfers.com.