The AI Premium in Hospitality M&A: Broker Story or Real Number?

I was reviewing this week’s deal sheet at 6:15 a.m. — coffee, two screens, the usual — when an associate at a mid-market sponsor sent me a one-liner: “Is the AI premium real or are we all just paying brokers to say it?” Fair question. I went back through the last twenty-four months of restaurant and restaurant-tech M&A, every transaction I could pin to a press release or an 8-K, and ran the numbers on the ones with disclosed multiples.

Here is the contrarian read up front: the AI premium, as brokers are pitching it on single-unit and small-group hospitality listings this spring, is mostly a story. The premium that is real is showing up one tier above the operator — in the platforms underneath the operator, where the buyers are public companies and PE funds who actually need to show their boards a software multiple. At the operator level, what’s being priced is distribution and data, not models. The brokers have just learned to spell the third word.

I’ll walk through the deals.

The data-layer grab: DoorDash buys SevenRooms

DoorDash announced the $1.2 billion all-cash acquisition of SevenRooms on May 6, 2025, and completed the deal on June 13, 2025. SevenRooms was running at roughly $150M ARR, which puts the deal at about 8× ARR. That is not a heroic SaaS multiple in absolute terms — median Q1 2025 SaaS M&A was 4.2× revenue with an average of 6.0× — but it’s a premium to median, and the strategic angle is unambiguous.

DoorDash did not buy SevenRooms for AI. They bought it for the CRM and guest data sitting across 13,000+ venues. Tony Xu’s framing in the press release was about enhancing the Commerce Platform — which is corporate-speak for “we want first-party guest data we don’t currently have.” The reservation flow is the wedge. The data is the asset. I wrote a longer Bottom Line on this thesis here, and our broader DoorDash commerce-platform read is here.

Buy-side read: 8× ARR is what you pay for a category-leading data set with a defensible enterprise customer base. The premium over median is the data, not the algorithm.

The distribution grab: DoorDash closes Deliveroo

Five months after the SevenRooms announcement, DoorDash closed the £2.9 billion (~$3.9B) acquisition of Deliveroo on October 2, 2025. The combined company now operates in 45 markets globally, 30 of them in Europe.

This is the trade that tells you what DoorDash actually thinks the premium is. Between the SevenRooms deal and the Deliveroo close, DoorDash committed roughly $5 billion in cash inside a six-month window. The Deliveroo number is more than three times the SevenRooms number. The Deliveroo asset is markets, riders, and consumer demand. The SevenRooms asset is software and data. If you want to know what DoorDash’s revealed preference is on the AI-vs-distribution question, look at the cheque size. Distribution wins by 3.25:1.

For context, DoorDash bought Wolt for $8.1B in an all-stock deal that closed May 31, 2022. Three large platform deals in three years — Wolt, SevenRooms, Deliveroo — and exactly one of them is a software company.

Buy-side read: when a strategic with the best AI talent in the category puts 75% of its M&A cash into geography rather than models, the platform-level “AI premium” is a story the sellers tell, not one the smartest buyers pay.

The take-private benchmark: Thoma Bravo buys Olo

This one’s the cleanest pricing comparable in the set. Thoma Bravo announced a $2.0 billion all-cash take-private of Olo on July 3, 2025, at $10.25/share — a 65% premium to the unaffected price of $6.20 on April 30, 2025. The deal closed September 12, 2025.

Olo did $285M in 2024 revenue, which puts the headline at 7× trailing revenue. On the Q1 2025 annualized run rate ($323M), it’s ~6.2×. That’s almost exactly the SaaS M&A average for the period. Olo had every “AI” tailwind brokers love to cite — restaurant tech, vertical SaaS, ordering rails into enterprise chains — and the public market priced it at $6.20 before the deal hit. Thoma Bravo paid a 65% control premium and got it for an average-of-market multiple.

Read carefully: the public market did not price an AI premium into Olo. A sophisticated software PE buyer paid the median, not the mean, after a control premium. If there were a real AI premium on enterprise restaurant tech, Olo is precisely where you’d see it. You don’t.

Buy-side read: this is the benchmark. Restaurant tech in 2025 traded at SaaS-median multiples, with the premium going to the seller for control, not for AI capability.

The card-network play: Amex / Tock (and the Resy backdrop)

American Express announced the $400M cash acquisition of Tock from Squarespace on June 21, 2024, and completed the deal on October 15, 2024. Tock had ~7,000 foodservice venues at close, which puts the price at roughly $57K per venue — a number that prices the customer relationship, not the software.

Amex already owned Resy (2019). Tock layers fine-dining and ticketed events on top. The thesis is naked: restaurant spend is one of the largest categories on Amex cards, and a card network that controls the reservation rails controls the front of the funnel. That’s not AI. That’s distribution geography rebuilt in software. I went deeper on the Resy/Amex strategic logic here.

Buy-side read: a card network paying for the booking layer at ~$57K/venue is paying for the cardholder funnel. Whatever AI Tock or Resy ships post-deal is operational efficiency for Amex, not the reason for the cheque.

The unglamorous distribution monster: Sysco / Jetro Restaurant Depot

Sysco announced the $29.1 billion acquisition of Jetro Restaurant Depot on March 30, 2026. Per Sysco IR, that’s 14.6× Restaurant Depot’s operating income — financed with $21B in new and hybrid debt plus $1B cash and equity, with shareholders also receiving 91.5M Sysco shares. Sysco stock dropped about 12% on the announcement — the market is not in love with the leverage.

This is the deal you have to put next to the DoorDash and Amex deals to see the pattern. Sysco — the largest broadline distributor in North America — paid 14.6× operating income for a cash-and-carry warehouse business. No AI in the deck. No software premium. The asset is independent restaurants buying ingredients in person at low margins. The strategic logic I wrote up at length here is about owning the price-conscious independent channel that Sysco’s truck-delivery model can’t economically reach.

If the AI premium were systemic, Sysco’s IR would have mentioned it. They didn’t. They talked about cash-and-carry as a higher-margin, resilient channel. The most expensive restaurant-adjacent deal of 2026 is, at its core, a real-estate-and-logistics trade.

Buy-side read: when the biggest M&A cheque in the category goes to warehouses and forklifts, the broker calling AI a “premium driver” on a 65-seat Italian deserves a hard look.

The tech-supplier roll-up: PAR / TASK (and Stuzo)

For completeness on the tech-supplier side: PAR Technology acquired TASK Group for $206M, announced March 2024, closing in Q3 2024, and acquired Stuzo for ~$190M in the same window, bringing total M&A spend to roughly $400M. The TASK consideration was $131.5M cash plus 2,163,393 PAR shares. TASK operates POS in 70 countries with clients including Starbucks, Guzman Y Gomez, and McDonald’s in 65 markets.

What did PAR buy? POS estate. International footprint. Enterprise logos. The pitch is “unified commerce platform” — front-of-house and back-of-house under one roof. There is no AI line item in the consideration. The premium is paid for installed base and chain relationships.

Buy-side read: enterprise restaurant tech consolidation in 2024-2025 is being priced on customer estate, not model capability. Same pattern as Olo. Same pattern as SevenRooms minus the data-layer narrative.

The pattern

Six deals, $36+ billion of disclosed transaction value, and one quietly consistent answer: buyers are paying for distribution, customer estate, and first-party data — in that order — and none of those is AI.

The closest thing to a real “AI premium” in this set is the SevenRooms multiple above SaaS median, and even there the asset is the guest data set, not the model. Olo, which had the strongest pure-software AI narrative of the group, traded at the SaaS median after a control premium. Sysco — the largest cheque — went to warehouses. Amex went to cards. DoorDash went to riders.

What the brokers are doing on indie listings — adding 15-20% to the asking multiple because the seller deployed an AI feature six months ago — has no support in the platform-level comparables. Operators who deploy AI tooling get real efficiency gains. They do not get a discrete premium at sale. They get the same multiple as their peers, hopefully on better EBITDA because the tools actually worked. That’s the right way to think about it: AI lifts the number you’re capitalizing, not the multiple you put on top of it.

If anyone tries to sell you a 5.5× single-unit multiple “because of the AI” — and I’m seeing this pitch weekly on the marketplace — point them at the Olo trade and ask why a pure-play restaurant SaaS with $285M of recurring revenue couldn’t clear 7× revenue. The honest answer is: the public market doesn’t believe in the premium either.

What an indie should do with this read this week

Three concrete actions:

1. If you’re considering selling in the next 18 months, stop building the AI-premium narrative into your CIM. Build the EBITDA narrative. Capable buyers will discount the AI line and capitalize the operating result. Sell the result. The brokers will not push back on this because it makes their job easier.

2. If you’re buying, demand the data-set diligence the platform buyers do. SevenRooms got priced for 13,000 venues of CRM history. Your target’s data set is smaller, but the same logic applies: how many months of clean POS, reservations, and inventory does the operator actually have? Anything less than 12 months is not an asset.

3. Watch the channel deals, not the model deals. Sysco/Restaurant Depot tells you where independent-restaurant economics are going (price-conscious, cash-and-carry, lower margin). Mise’s four-margin frame is the operator’s playbook for that environment. If you’re not pricing the four margins explicitly, the broker selling you “AI-enabled” is wasting both your times.

I’ll be back Monday with the casual-dining quiet list — operators not currently selling who are starting to talk like they might. The DoorDash cheque book has not run dry; the AI premium narrative is still circulating on the sell-side; and somebody’s about to overpay. Tips before the brokers see them: tips@tabletransfers.com.

— Marcus runs The Bottom Line and gets the deal flow before the brokers. Tips: tips@tabletransfers.com.