The ghost kitchen hype cycle is over. Most of the venture-funded mega-brands folded, the "just list on five apps" playbook stopped paying, and what's left in 2026 is a model that can throw off a healthy margin or lose money for a year before anyone notices. Which way it goes comes down to a variable most coverage skips: who owns the order, not what's on the menu.
What the model genuinely saves
A ghost kitchen cuts the two heaviest lines in a normal restaurant P&L. Rent drops because you're renting a cooking box in a cheap location instead of a dining room on a visible corner; operators put occupancy at roughly half of a comparable sit-down spot. Labor drops further. There are no servers, hosts, or bussers, so front-of-house payroll never lands on the sheet.
Add a faster launch (weeks, not the year a full build-out eats) and the option to run several menus off one line, and the appeal is obvious. On paper the model starts 15–20 points of margin ahead before a single order is cooked.
The less obvious advantage is speed of iteration. A dining-room restaurant reworks its concept by renovating; a ghost kitchen does it by editing a menu and reshooting the photos. A bad idea dies in a month for the price of some packaging instead of a year of lease payments. For operators who treat the kitchen as a test lab, that's worth about as much as the rent savings.
Where the savings leak back out
Now the other side of the ledger. A ghost kitchen has no storefront, no walk-ins, no sign in a window. Every customer has to arrive through a screen, and when that screen belongs to a delivery marketplace the economics flip. Published commission tiers run 15%, 25%, and 30% — DoorDash lists Basic, Plus, and Premier at exactly those levels, and Uber Eats covers the same 15–30% (rates raised March 2026). Add promos, processing, and refund charge-backs, and the real cost lands at 30–40% of every order. We ran the line-by-line math separately.
So do the arithmetic. You saved 15–20 points on rent and labor, then handed 30–40 points of revenue to the marketplace. A ghost kitchen that pulls all its volume from delivery apps ends up structurally worse off than the dining room it replaced, and with no channel of its own to fall back on. That math, more than any complaint about food quality, is what killed so many ghost brands between 2022 and 2025.
We'll model your channel mix — marketplace vs. direct — on a 30-minute call, with your ticket sizes.
The version of the model that works in 2026
What the survivors have in common is a channel structure, not a menu style. They use marketplaces for the one thing marketplaces do well, which is discovery and first orders from strangers, and route repeat volume through a channel they own: a branded ordering app and website where the cost per order is payment processing rather than 30%-plus. The conversion numbers back this up. Across 3M+ orders on our platform, branded apps convert up to 35% of visitors and ordering sites up to 15%, against roughly 3% on a marketplace listing. Once a customer installs your app, the marketplace can't bill you for them again.
They also run tight on space. With no dining room to absorb slack, the kitchen is the entire business, and it has to be organized like one. Orders from every channel land on one order-management screen instead of five tablets. Dispatch and courier tracking run without someone routing by hand. And once a week somebody actually reads item-level margins in the reports, because in a ghost kitchen a dish that loses money has nowhere to hide.
And they multiply. One kitchen can carry several virtual brands: wings under one name, bowls under another, sharing the same fridge and the same payroll. That lifts revenue per square meter without lifting fixed cost, which is the whole lever the model exists to pull. We cover the mechanics in the virtual brands article.
Who should open one, and who shouldn't
It works best for operators who already have a delivery-proven menu and a customer base to seed the direct channel. Multi-brand operators fit too, since they can spread one kitchen across two or three concepts. So do established restaurants that want to test a new neighborhood with a ghost unit rather than a full build-out.
It punishes two kinds of operator. First-timers with no audience inherit the marketplace's 30–40% effective take and have no repeat channel to offset it. And premium dine-in concepts don't translate, because the value there is the experience, which doesn't survive the trip in a paper bag.
What the box, the equipment, and the stack actually cost is its own subject, and we broke that down line by line here.
Before you sign a lease
Before you sign anything, write down the channel mix you're assuming: what share of orders comes through marketplaces, what share through your own channel, and the effective cost of each. If the blended platform cost is higher than what you save on rent and labor, the model doesn't work at that mix. Fix the mix on paper before you fix it in a lease, and build the direct channel into the launch plan instead of leaving it for year two.
App, ordering site, order management, dispatch, and marketing — the owned half of the ghost kitchen model, from one platform.