The DoorDash Decade: Why May 2025 Is When Food Delivery Became Hospitality Infrastructure

It is the last Friday in May, my desk is mostly paper, and the coffee is at that point in the afternoon where it has stopped being warm and started being a prop. I have three printouts in front of me: the DoorDash press release announcing the SevenRooms acquisition, the BusinessWire copy of the Deliveroo agreement, and DoorDash’s Q1 financial results. Next to them, a notebook page where I have been drawing the same diagram for two weeks — a square divided into four quadrants labeled delivery, reservations, CRM, and payments — and writing the same word above it in slightly different handwriting each time: stack.

This is the May Mise. I write one of these a month, and the job is not to recap what happened — by the time I sit down to write the synthesis piece, the news cycle on a given month’s big events is three weeks old and the operators I respect have already had the meeting. The job is to put a frame around the month that the operator can carry into June, and into July, and ideally into the conversation they have with their next vendor.

So here is the frame. On May 6, 2025, Tony Xu spent approximately $5 billion in a single day — $1.2 billion in cash for SevenRooms and £2.9 billion (about $3.9 billion) for Deliveroo — to turn DoorDash into the hospitality industry’s first vertically integrated stack. Delivery, reservations, CRM, payments, in one corporate entity, with one balance sheet, one product roadmap, and one go-to-market motion. That is the structural event of the month. Everything else that happened in restaurant tech in May — every funding round, every press release about an AI ordering bot, every NRA Show post-mortem — is a footnote to it.

The Mise take, which is the only take I am willing to defend at length, is this: the next decade of restaurant tech will be defined by the consolidation Tony Xu’s May 6 announcement triggered, and operators who do not reframe their vendor portfolio as “one stack versus best-of-breed” within the next twelve months will be making the decision retroactively, on terms set by someone else. That is the framework. The rest of this essay is me defending it.

What May 6 actually consolidated

I want to be precise about what was bought, because the casual read of the headlines — “DoorDash buys reservation company, DoorDash buys European delivery company” — undersells the structural shift by an order of magnitude.

SevenRooms is not a reservation app. It is a guest data platform with reservations as the wedge. The product serves more than 13,000 venues, and the actual asset is the structured guest record — the dietary preference flagged on the third visit, the wine the table sent back on the fifth, the spend pattern that says this two-top is worth pulling forward off the wait. The reservation flow is how the data gets into the system. The CRM and marketing automation are how the data gets back out into revenue. PSG, which had owned a minority stake in SevenRooms, exited at the close. The CNBC writeup at cnbc.com/2025/05/06/doordash-announces-1point2-billion-sevenrooms-deal-misses-revenue-expectations.html frames it as a reservation deal because that is the legible piece; the SevenRooms press release at sevenrooms.com/press/DoorDash-Acquires-SevenRooms/ is more honest about what is changing hands, which is the guest graph.

Deliveroo is the geographic half. The £2.9 billion agreement, announced via BusinessWire at businesswire.com/news/home/20250505418218/en/DoorDash-Announces-Agreement-to-Acquire-Deliveroo, gives DoorDash the European and Middle Eastern footprint it has been triangulating toward through Wolt for two years. Bloomberg’s writeup at bloomberg.com/news/articles/2025-05-06/doordash-buys-sevenrooms-for-1-2-billion-in-restaurant-tech-bet is the one that put both deals on the same page, which is how they should always have been read — not as two transactions but as one strategic posture announced in two press releases on the same morning.

The frame I keep coming back to is what is not in those press releases. There is no language about a reservation product feature roadmap. There is no language about a delivery integration milestone. The language is about platform, footprint, guest, and stack. Tony Xu’s commentary on the Q1 earnings call, archived at ir.doordash.com/news/news-details/2025/DoorDash-Releases-First-Quarter-2025-Financial-Results/default.aspx, treats global expansion and the SevenRooms deal as two expressions of the same thesis: DoorDash is no longer building a delivery network with adjacent products bolted on. It is building a hospitality operating system, and delivery happens to be the surface area through which the operator first met it.

That is what May 6 consolidated. Not two product lines. One thesis.

Why the SevenRooms half is the operator story

If you run restaurants, the Deliveroo half is interesting and the SevenRooms half is consequential.

Here is why. For most of the last decade, the operator’s stack has looked like a layered cake of point solutions, each chosen on its own merits in its own quarter: a POS, a reservation system, a marketing/CRM tool, a delivery integration, a labor product, an inventory product, a gift card vendor. Each one was best-of-breed in its layer. Each one was integrated to the next via some combination of native API, middleware, and operations-manager spreadsheet. The bet, implicit in every one of those choices, was that the marginal benefit of picking the best tool per layer outweighed the marginal cost of running the integrations.

That bet was rational when no single vendor could credibly offer a competitive product in more than two layers. SevenRooms-plus-DoorDash changes the math because the combined entity now has competitive products in delivery, reservations, CRM, marketing automation, payments (via the existing DoorDash payments work), and — through the data that flows across all of those — a guest record that none of the layer-specialists can replicate.

The operator question that follows is not “should I switch to the DoorDash stack?” The operator question is: how much value does my current best-of-breed configuration leave on the table relative to a stack where the guest record is unified across reservations, delivery, marketing, and payments?

That is a harder question than it looks, because the answer depends on the operator’s gross margin profile, the share of revenue that comes through each channel, and the degree to which the operator’s existing CRM is actually being used. For a high-volume, delivery-heavy concept where the marketing CRM is a license that nobody opens, the gap is enormous. For a fine-dining group with a head of guest experience who lives inside SevenRooms already, the gap is smaller but the direction of the gap is the same: the unified stack now has a structural advantage that compounds with every additional channel the operator runs through it.

That is the operator story. It is not “DoorDash bought a reservation company.” It is “the dominant guest-data platform in the segment that has the most operator-loved data product just merged with the dominant delivery platform in the segment that has the most operator-leveraged transaction product.” The vertical is real.

Why the Deliveroo half is the geographic story

The Deliveroo half is the part that gets the headlines in Europe and the shrug in the U.S., and that is exactly backwards from how the operator should read it.

The geographic argument is straightforward. DoorDash’s Wolt acquisition gave it the Nordics, parts of Central Europe, and a Tel Aviv-to-Tokyo arc that filled in the map but did not give it the U.K. or the Middle East at the density required to matter. Deliveroo does. Foodondemand’s follow-up piece at foodondemand.com/05222025/what-next-for-doordash-after-acquiring-sevenrooms/ does the math on the combined footprint and the answer is “DoorDash is now a credible global hospitality platform in a way that, before May 6, it was not.”

The operator-side argument is the one I think gets missed. If you are a multi-unit operator with U.K. or European exposure — and that is a much larger group than the U.S.-only conversation tends to assume, because the chains that drive the segment-level economics increasingly have transatlantic footprints — then May 6 is the day the vendor map for your European operations got redrawn to match the vendor map for your U.S. operations. One contract. One data model. One reconciliation flow. That is not a minor convenience. For groups in the 50-to-500 unit range with mixed U.S./European footprints, it is the difference between running two finance stacks and running one.

The geographic half is also what funds the operator-side ambition. DoorDash’s ability to invest in the SevenRooms product roadmap, to build the integrations that make the unified stack genuinely better than the sum of its parts, depends on the cash generation of the delivery business at global scale. Deliveroo is the cash; SevenRooms is the moat.

The Q1 capital base that funded both

I want to spend a paragraph on the Q1 numbers because they are the part of the May 6 announcement that explains why this deal happened in May 2025 rather than in May 2024 or May 2026.

DoorDash’s Q1 revenue was $3.0 billion, up 21% year over year. Adjusted EBITDA was $590 million. Free cash flow for the quarter was $494 million. Those are not the numbers of a company that bought SevenRooms for $1.2 billion in cash because it had to stretch; they are the numbers of a company that bought SevenRooms in cash because it could, and because doing it in cash signals to the market and to future acquisition targets that the next deal can also be in cash. The Deliveroo transaction is structured differently — it is a recommended offer at the £2.9 billion headline — but the same underlying point applies: a company generating roughly half a billion in free cash flow per quarter can announce a $5 billion combined transaction and have it read as confident rather than overextended.

The principle here, and this is where the Mise frame gets to do some work, is that the consolidation phase of restaurant tech requires an acquirer with a balance sheet, and as of Q1 2025, DoorDash is the only pure-play restaurant-tech-adjacent company with that balance sheet. Toast is bigger by some measures but cash-generative on a different timeline. Olo is a specialist. The international players are regional. The capital base that funded May 6 is the same capital base that will fund the next three deals, and the operators reading this should be thinking about which point solutions in their current stack are the most likely next targets.

What “one stack” means at the operator level

I want to be concrete about what the unified stack actually does, because the abstraction “vertical integration” has been used to sell a lot of things that turned out to be a single login over a fragmented backend.

A real unified stack, as I am defining it for the purposes of this framework, is one in which:

The guest record is the same object across reservations, delivery, marketing, and payments. The diner who books a four-top through SevenRooms is the same diner who orders delivery through DoorDash three weeks later, and the spend, frequency, dietary, and preference data flow into the same record without an integration layer in between.

The marketing automation can target across channels with a unified attribution model. The operator who wants to win back a lapsed delivery customer with a dining-room offer, or pull a high-LTV regular forward with a delivery promotion during the slow Tuesday window, can do it from one product and measure it with one set of numbers.

The payments rails are consolidated. The reconciliation that today takes an operations manager four hours a week, because the delivery payouts and the dining-room settlement and the gift-card redemption come from three different processors with three different reporting cadences, becomes one statement.

The reporting layer is unified. Channel mix, daypart mix, guest cohort behavior, marketing ROI by campaign — all of it sits on the same data warehouse, and the operator does not have to be the ETL pipeline.

That is what the May 6 transaction sets up as the medium-term product roadmap. None of those four things is delivered today on day one of the post-close integration. All four of them are credible eighteen-to-thirty-six-month deliverables given the assets DoorDash now owns and the capital base it can deploy to integrate them. The operator who is making vendor decisions in the next twelve months should be making them with the assumption that the unified stack will be real by 2027 and reasoning about what that means for switching costs accrued between now and then.

Where best-of-breed still wins

I am not going to write a framework essay arguing for consolidation without acknowledging the cases where the consolidation thesis loses, because operators read this column and the ones I respect will throw it across the office if I do not.

Best-of-breed wins in three places.

The first is at the high end of the guest experience, where the marginal value of a deeply specialized reservation, host, and CRM workflow exceeds the marginal value of cross-channel data unification. A three-star tasting menu does not benefit much from knowing what the guest’s delivery order history looks like, because the guest does not have one, and the host team’s needs are specialized enough that a generalist platform will lag the specialist for years. SevenRooms inside DoorDash will keep that segment, but the next SevenRooms — the next specialist that out-features the unified stack on the guest workflow — will find a market.

The second is at the deeply operations-heavy middle, where the POS-and-back-of-house workflow is doing more of the work than the guest-front-of-house workflow. Toast’s segment, broadly. The unified stack DoorDash is building is a guest-and-channel stack, not a kitchen-and-labor stack, and operators whose margin is being made or lost in the kitchen will continue to optimize the kitchen-side vendor independently. The two stacks will integrate, eventually, but they will not merge.

The third is in markets where DoorDash does not yet have density, even after Deliveroo. Parts of Asia, parts of Latin America, parts of the Middle East outside the Deliveroo footprint. The local champion in each of those markets will continue to be the right vendor for operators whose center of gravity is local, and the global consolidation argument will reach those markets on a delay of three to seven years.

The principle is that the unified-stack thesis is dominant in the segments where the guest record is the operator’s primary unit of value creation, and weaker in the segments where the kitchen or the local market is. That is most of the segments that read this column. It is not all of them.

The 12-month decision window

Here is the part of the framework that I want operators to actually act on, because most framework essays describe a structural shift and then leave the reader without a decision to make, and that is the part of the genre I am trying to fix.

The next twelve months are the decision window. Between May 30, 2025, and roughly the end of Q2 2026, every multi-unit operator with more than a hundred covers per day across more than three locations is going to be making one of three decisions about their stack, whether they realize it or not.

Decision one is to commit to the unified stack early. Sign with the DoorDash side of the consolidating vendor, accept that the product roadmap will be imperfect for the first eighteen months while the integrations get built, and lock in the long-term switching-cost advantage of being on the platform when the data unification work is finished. The operators who take this path will be the case studies the vendor uses in 2027, which is its own form of leverage on commercial terms.

Decision two is to commit to best-of-breed deliberately, with a clear-eyed view of where the unified stack would beat the layered configuration and a willingness to invest in the integration work that holds the layered configuration together. This is the right answer for the high-end and the kitchen-heavy segments described above. It is the wrong answer dressed up as the right answer for the broad middle, which is where most operators sit.

Decision three is to not decide, which means continuing to renew each layer on its current contract and letting the vendor map drift. This is the path most operators will take, because vendor decisions are exhausting and the day-to-day pressure of running restaurants does not leave a lot of room for portfolio-level strategy. The cost of decision three is that twelve to eighteen months from now, the operator will discover that the unified stack has produced a meaningful product gap with their current configuration, and the switching cost at that point will be higher than it would have been if they had decided in May 2025.

What I am holding as the operator-facing recommendation is decision two for the segments where it genuinely wins, decision one for the broad middle, and a hard internal commitment to not letting decision three happen by default. The way to avoid decision three is to put the stack question on the agenda of one specific meeting between now and the end of Q3, with the vendor map drawn out on a whiteboard and an explicit yes-or-no on whether the layered configuration is still the right answer given May 6. That is the meeting. If you have it, you have made a decision. If you don’t have it, May 6 made the decision for you.

What the precedents say

Whenever a vertical-integration thesis lands in a market, the right discipline is to look at the prior verticalizations in adjacent markets and ask which precedent is the right one to reason from.

There are three that I have been working with.

The travel-tech precedent — the consolidation of OTAs through the 2000s and 2010s — is the one most often invoked, and it is the wrong precedent. Travel verticalized around inventory aggregation, not around the guest record, and the unit of value is the booking rather than the relationship. Restaurants are the other way around.

The retail e-commerce precedent — the Amazon stack, where logistics, payments, advertising, and merchant tooling consolidated under one roof — is closer. The dynamic where the platform becomes the demand source, the data source, and the operations layer simultaneously is the dynamic DoorDash is now positioned to replicate in restaurants. The thing to watch from that precedent is the relationship between the platform and the merchant, which over a decade evolved from partnership to dependency to active rebellion in some segments. That arc is the one the restaurant operator should be most attentive to, because it tells you what the endgame looks like if you let the platform accumulate the relationship with your guest.

The healthcare-tech precedent — the consolidation of EHR, scheduling, and revenue cycle management under Epic and a handful of others — is the most structurally similar and the one I think reasons best. In healthcare, the stack consolidated around the patient record, and the operators (hospitals, practices) ended up with a small number of platform vendors whose products were not the best in any individual layer but whose unified data model was indispensable. The specialists survived in niches; the broad middle moved to the stack. That is the future I think we are looking at in restaurants, and the May 6 transaction is the day the analog of Epic-buys-MyChart-buys-the-billing-platform happened in our market.

The principle is that the unit around which a market verticalizes determines who wins, and in restaurants the unit is the guest record. DoorDash, as of May 6, owns the path to the most complete guest record in the industry.

What this sets up for June’s operator-vendor compact essay

The framework chain matters, and the next link in the chain is the question this essay’s thesis raises but does not answer.

When one vendor owns delivery, reservations, CRM, marketing automation, and a significant share of payments, what does the operator-vendor relationship actually become? It is no longer the procurement-and-integration relationship of the layered-cake era. It is something closer to a long-term commercial partnership in which the operator’s data is the joint asset, the vendor’s product roadmap is part of the operator’s strategic plan, and the contract has to do a lot more work than the standard SaaS agreement was ever designed to do.

That is the essay I will write at the end of June. The working title is the operator-vendor compact, and the question it has to answer is: given that the unified stack is a structural reality, what is the right shape of the commercial and data-sharing arrangement between the operator and the consolidating platform? What do operators need to negotiate for that they have not historically negotiated for? What are the data-portability terms that protect against the lock-in risk? What is the right governance structure for a relationship that is now strategic rather than transactional?

I am not going to preview the answer because I have not finished thinking it through. What I will say, as the framework-chain link, is that the May Mise — this essay — is the structural diagnosis, and the June Mise is the response. You cannot reason about the right operator-vendor compact until you have accepted that the vertical integration is real and dominant. That is what May 6 did, and that is what this essay is for. June is the next move.

For readers tracking the longer framework arc, the framework piece we develop further later in the year — the four-margins-of-a-restaurant essay at /blog/posts/the-four-margins-of-a-restaurant — picks this thread up again from the margin side, asking which of an operator’s four margins is most affected by the vertical consolidation described here. And the framework we publish later building on this essay’s verticalization frame — the voice-agent maturity curve revisit at /blog/posts/the-voice-agent-maturity-curve — extends the stack question into the AI-mediated guest interface layer, which is the next vertical to consolidate after the one we are documenting this month.

What I am holding through year-end

Every Mise ends with a list of the positions I am holding through the rest of the year, partly so that future-me can be held accountable and partly because the most useful thing a framework essay can give an operator is a short list of things the writer is willing to be wrong about in public.

I am holding the view that vertical integration of delivery, reservations, CRM, and payments is the dominant structural shift in restaurant tech, and that the May 6 transaction is the inflection point at which the thesis became undeniable rather than speculative. If by year-end the integration work has stalled or the combined entity has retreated from one of the four legs, I will write that essay.

I am holding the view that best-of-breed survives in the high-end guest segment and in the kitchen-heavy operations segment, and not much elsewhere in the broad middle. If by year-end a credible best-of-breed counter-stack has emerged with a real product unification story, I will revisit.

I am holding the view that the next twelve months are a real decision window for operators, and that decision three — the no-decision path — is the most expensive of the three options on a two-to-three-year horizon. If by year-end the unified stack’s product progress has been slower than expected and the cost of waiting has gone down, I will revise the urgency.

I am holding the view that the next consolidation moves come from the same balance sheet that funded May 6, and that operators should be reading the Q2 and Q3 DoorDash earnings calls with an eye for which adjacent vendor category is being signaled as the next target. The payments layer is the one I am watching most closely.

And I am holding the view that the June Mise on the operator-vendor compact is the more important essay than this one, because diagnosing the structural shift is the easier half of the job and prescribing the operator’s response is the harder half. This essay sets the table. The next one is the meal.

It is six in the evening now. The coffee is cold and the diagram in the notebook has had the word stack underlined three times. I will close the notebook, file this, and start the June piece tomorrow. The framework is the thing I owe you every month. May 6 made my job easier this time, because the structural event was unambiguous. June will be harder, because the response is.

That is the May Mise. See you at the end of June.

— Eitan writes the Mise column. Tips: tips@tabletransfers.com.