The Second Re-Rating

Since the labor market reset after 2020, every operating meeting in this business has been about cost. Sharper forecasting, tighter standards, cross-training, scheduling indexed to booking pace. It is the correct conversation to be having about the P&L, and it is the wrong one to be having about artificial intelligence.

The efficiency case has been made repeatedly, mostly by people selling SaaS, and it is the smaller half of the story. What interests us is the multiple.

Hotels trade at wider capitalization rates than any other institutional real estate class. Not because the buildings are inferior, but because a hotel is not really a building — it is an operating business that reprices its entire inventory every night and carries a payroll to do it. The spread against multifamily, against industrial, against net lease is the market’s standing charge for operational intensity. Labor is its largest single component.

That charge is not permanent. It is a price attached to a constraint, and the constraint is moving.

The market has done this before

Select-service did not become the preferred institutional format by accident. Hampton, Holiday Inn Express, Fairfield and their peers command strong pricing precisely because they removed the food and beverage complexity that made full-service difficult to operate and harder to underwrite. Investors did not penalize the simpler product. They paid up for it.

That was a re-rating, and it happened within the professional memory of most people reading this. An entire category was repriced — not because its buildings improved, but because its operating burden fell and the market recognized the earnings underneath as better and more predictable.

Artificial intelligence is the next simplification, and it reaches something select-service never touched: the cognitive labor that remained necessary whether or not a kitchen was attached.

What the numbers say

According to HotelData.com’s 2025 labor report, drawn from thousands of U.S. properties, wage cost per occupied room rose 12.8 percent last year, from $42.82 to $48.32. Taken alone that proves little; wages have risen everywhere since 2020, and much of what a hotel now pays per hour is simply inflation carried forward and never given back.

The second figure is harder to explain away. Over the same period, hours per occupied room rose 4.4 percent.

Inflation raises what an hour costs. It does not require more hours to clean the same room. Operators paid more per hour and consumed more of them — in the year they were running their most disciplined scheduling since the pandemic. Fourth-quarter gross operating margin compressed 3.3 points to 36.0 percent.

The response has been to manage harder: sharper forecasting, tighter standards, labor tools indexed to booking pace. It is sound practice, it has been executed well across the recovery, and the cost line has outrun the revenue line throughout. When competent management fails to arrest a trend, the trend is not an execution problem.

Two technologies, one of them new

The industry discusses this as one thing. It is two.

Mechanization is not new. Machines have been removing physical repetition from the back of house for forty years — tunnel washers, automated folding, floor equipment. It advances with hardware cost, it addresses work that is dull and identical, and no guest has ever formed an impression of a hotel from the far side of a laundry door.

Artificial intelligence is the new variable, and what it addresses is not physical labor but cognitive labor — understanding, resolving, deciding. Answering an ordinary question in ordinary language. Reading a document. Reconciling a ledger. Handling the exception that does not match the script.

That was always the untouchable half of a hotel’s payroll. It could not be mechanized because it was not mechanical. It required a person, and so a person sat there — often at three in the morning, often for the sole reason that no alternative existed.

That is the constraint that broke. Not the cost of machinery, which was falling anyway, but the sudden addressability of work requiring comprehension.

The objection

There is one, and it is the right one: hospitality is a human business, and an operation stripped of people may be efficient and not worth paying for. It is the industry’s founding article of faith and for most of its history it was simply correct.

Whether it remains correct is a question about the guest rather than about the technology, and it is large enough to deserve its own entry. We take it up next. For the present note it is sufficient to say that the answer is no longer obvious, that guest behavior has moved further than the industry’s assumptions about it, and that we do not think the objection survives contact with what people actually do.

The re-rating

Here the operating story becomes an investment story, and here most commentary stops. The industry treats artificial intelligence as an efficiency program — a line-item saving assessed on payback period, adjacent to an LED retrofit or a laundry contract. That framing understates it by an order of magnitude, because it prices AI as a vendor expense rather than as what it actually is: a permanent change in how much labor a hotel of a given size requires to operate.

A property carrying a structurally lighter labor load is not the same asset as the one across the street flying the same flag at the same rate. It converts more of every revenue dollar, and it does so by construction rather than by effort. A buyer cannot replicate it by managing more attentively.

That is not a cash-flow adjustment. It is multiple expansion.

We expect the market to reward this the way it rewarded the last simplification — not by paying more for the incremental income, which is ordinary capitalization, but by assigning a different multiple to earnings it now regards as structurally better and more predictable. The re-rating appears in price per key and in revenue multiple, on the same building, on the same flag, at the same rate.

That is the paradigm shift. Not that hotels will cost less to run — that part is arithmetic, and it is being discussed to exhaustion. That an asset class carrying a permanent discount for operational intensity gets repriced as the intensity comes out of it.

Those positioned early capture the expansion. The remainder inherit the arithmetic.

We underwrite on the premise that the operating recovery available to a disciplined operator exceeds what the incumbent can execute. That spread is where our returns have always originated. What has changed is its width — and, for the first time, that some portion of it is structural rather than a matter of effort.

We will be wrong about parts of this. We will say so here.

Every operating standard in this business was once an extravagance and before that an impossibility. The sequence never varies: novel, then differentiating, then expected, then invisible. Capital positioned against a secular shift is not protected by conviction — it is marked to a market that moved without it.

← AI, Live