A fleet that was beating its forecasts got scrapped: nobody would fund one kitchen per customer

The product

Wonder is an app that takes one order across more than twenty restaurants and brings it to the door as a single delivery. In its own words it is "the food delivery platform to feed every craving, all in one order". The restaurants – Tejas Barbecue out of Tomball, Texas, Mr. D's fried chicken out of Lorman, Mississippi, Di Fara pizza from Brooklyn, a steak under Bobby Flay's name, a kids' menu called Bellies – are exclusive to it: the recipes were bought or licensed, and every one of them is cooked in Wonder's own kitchens. There were 140 of those along the East Coast in July 2026, each one a shop with a counter and a few tables: delivery is free, pickup and dine-in happen in the same room, and when Restaurant Business visited the Upper West Side in September 2023 about 5% of guests were ordering in person.

How it works

The cooking is split in two. Proteins are cooked sous vide by the tub-full at a plant in New Jersey, pizza dough is parbaked, sauces are mixed; in 2023 that part-cooked food went out through nine cross-stops to the stores, where it sits refrigerated until somebody orders it. The finishing happens in the shop, on three pieces of equipment – TurboChef ovens, water baths and self-venting fryers – under what Lore calls edge cooking, and the company says the staff job "amounts to pushing a button on the oven". Wonder's phrase for the deal with the chefs is that they "teach us how to replicate their dishes exactly", after months of testing and a final seal of approval; the steak that started it took two months of trials before the smoke got the chef thrown out of a rented office, and two more years to get right.

A single order can mix as many of those menus as the customer likes, timed so the dishes finish together, and a Wonder courier – a uniformed employee, not a platform driver – carries it to an address within a six-minute drive or e-bike ride. The company's stated average from order to first bite is under thirty minutes. The same app also lists hundreds of ordinary neighbourhood restaurants, delivered by the same couriers at no delivery fee; those cannot be mixed into an order with Wonder's own menus.

Business model

Wonder owns the brands it sells. Independent restaurants that joined "get cash and stock in Wonder, and Wonder gets to use their brands and recipes without ever paying a cent in royalties", which Lore says "makes our financial model way better"; the celebrity chefs are paid the same way, in up-front money and equity, so that, as its marketing chief puts it, "they are incentivized to encourage the success of Wonder as a whole". There is no franchisee, no landlord tenant, no third-party courier: in January 2023 Restaurant Business summed the design up as every penny the customer spends going to Wonder.

The customer pays menu prices, nothing for delivery, and a service fee that varies with the order total – removed, along with priority in the kitchen queue and 5% back on pickup, by a $7.99-a-month membership the company says saves an average of $6 an order. Around that sit the businesses Wonder bought: Grubhub, a marketplace of more than 375,000 merchants and 200,000 delivery partners whose restaurants now appear inside the Wonder app while Wonder's kitchens appear inside Grubhub's; Blue Apron's meal-kit subscriptions; Tastemade's video audience. In September 2026 it agreed to sell the one piece that feeds none of its kitchens, Grubhub's campus dining business, to DoorDash for $300M, with the deal due to close in the first half of 2027.

The spike

Wonder's first kitchens had wheels. From 2020 it served Westfield, New Jersey, out of Mercedes-Benz Sprinter vans fitted with small ovens, and the promise on its homepage as late as December 2022 was exactly that: "Your order is fired, finished, and plated just steps away from your door, then served as soon as it's ready." Six months after launch there were 19 mobile restaurants covering 22 towns and more than 130,000 households, and in June 2022 Bain Capital Ventures led a $350M round at a $3.5B valuation to go on building them.

On 10 January 2023 Lore wrote that the company was "reducing the number of mobile restaurants we have in-market, and replacing them with physical locations" – delivery, pickup and a few tables, all the cuisines out of one room. His stated reason was what a pilot store had shown: "even faster order-to-eat times, higher NPS scores and cost-savings (through less overhead and infrastructure)", at the price of "tough personnel choices". Several hundred Sprinters were sold, the mobile operation was shut by the end of May and more than 100 people lost their jobs – and by his own account in that post the concept he was dropping was working, with revenue per van beyond what he had projected years earlier.

Two fuller explanations have followed, and they do not say the same thing. To Restaurant Business in September 2023 the problem was physical: "Parking 450 vans was not only expensive, but it also presented a barrier to future growth. Wherever Wonder went, it would have to find parking, and lots of it." To Fortune in 2025 it was financial. Investors had been asking how big a suburban van business could get and how long a fleet of kitchens on wheels would take to pay back; more than 100 pitches went out and, as Fortune puts it, the story just wasn't selling. A single storefront test was a hit within weeks, and Lore decided: "I had proven to myself that this was a superior model, and it had a superior story that was easier to raise money off the back of."

Read together, the two accounts describe one arithmetic. A van is a kitchen let to one household for the length of one order, and thirty cuisines in a neighbourhood meant thirty vehicles that had to be parked somewhere; a 2,800 square-foot room does the same thirty menus for a whole town, and lets one order cross several of them, which the vans never could. What Wonder dropped was not its product but the unit that carried its cost – and it dropped it while that unit was beating plan, because the plan was not the constraint. The money was.

What the vans left behind is the company's method: how to produce restaurant food fast, in a small space, with lightly trained staff. The menus, the app, the couriers and the sous-vide plant all crossed over unchanged.

Then a second turn, less announced than executed. From late 2023 Wonder stopped building demand and started buying it, always from owners who had given up: Blue Apron for $103M, a company valued at nearly $2B after it went public; Grubhub for an enterprise value of $650M, three years after Just Eat Takeaway.com had paid about $7.3B for it and years after it had carried a $20B market capitalisation; Tastemade's audience for about $90M; and Sweetgreen's Infinite Kitchen robots, a business Sweetgreen had bought for about $70M, for $100M in cash and $86.4M in Wonder stock. Stord ran the same move in the other direction – a marketplace that bought the warehouses it had been brokering; Wonder is the operator that bought the marketplace.

Why it matters

The alternatives cost nothing to install and were already on every phone; Grubhub alone listed more than 375,000 merchants. What they left to the household was the choosing and the consequences of the trip. Lore's diagnosis was that "food delivery was exploding, but people weren't happy with the quality of the experience"; Fortune later described Wonder's bet as removing "one of the last frictions left: agreeing on where to order from". Wonder's answer was to own both ends – its own menus, its own ovens, its own couriers, a delivery radius of about six minutes – and to let one order cross as many kitchens as it likes.

People used the part they could not get elsewhere. At the Westfield opening in June 2023, nearly half of the first Sunday's orders included items from more than one restaurant. By March 2024 Lore was claiming an under-30-minute average order-to-eat time, a Net Promoter Score above 60 and unit-level profitability, all of it self-reported; on that basis the estate went from 11 locations to 46 by May 2025 and 140 by July 2026, roughly one opening a week. The weak spot is the thing the chefs were bought for: Fortune found Google ratings for Wonder locations ranging from 2.6 to 4.6 stars, which for a company selling consistency is the number to watch.

The playbook

Owning every brand on the menu, with recipes paid for in cash and equity rather than royalties, lets one kitchen sell what an aggregator cannot: one order across many restaurants, delivered free. A household faces the same gap wherever it must agree on one supplier per purchase while marketplaces keep sellers apart. At Wonder's Westfield opening, nearly half the first Sunday's orders drew on more than one restaurant. The risk is consistency: Google ratings run from 2.6 to 4.6 stars.

The lesson

Wonder's costliest thing was a van, and a van cooks for one household at a time. Thirty cuisines in a neighbourhood meant thirty vehicles; 450 of them made parking the limit on growth, and every extra customer in the same hour needed another van and another driver. Nothing the company sold changed when it moved into rooms – same menus, same app, same couriers – only the unit carrying the cost, which went from one customer to one town. If serving one more customer at the same moment means buying another copy of your most expensive asset, capital sets your ceiling long before demand does: keep what you sell and rebuild it on an asset that serves a whole area at once.

That is paid for twice. The customer now comes to the asset or waits for it, so the area must be dense enough to fill the room, which is a lease and a payroll before it has an order. And the move is made against evidence: the fleet was scrapped while revenue per vehicle beat forecast, because a thing that works and a thing that can be financed are different tests. Which of your costs buys capacity for exactly one customer at a time, and what would it take to serve a hundred with it?

The second decision shows in what Wonder bought next: a meal-kit company, a marketplace, a media audience and a robot maker, each from an owner who had written it down, each pointed at kitchens already built. Demand someone else paid to create is cheaper than demand you create, and cheapest when its owner has stopped believing in it – after which the parts that feed nothing of yours are sold on, as the campus business is. Which company near you is worth less to its owner than its customers would be worth filling your capacity, and what did that owner pay for them?