Ghost kitchens in 2026: does the delivery-only model still work?

A few years ago, delivery-only kitchens were pitched as the future of eating out, restaurants with no dining room, no signage, no waiters, just a stove and a stream of app orders. Then the model collided with its own math. Here is what actually happened, and where it still holds up.

What a ghost kitchen actually is

A ghost kitchen, also called a cloud kitchen, dark kitchen, or virtual restaurant, is a food-production operation built to fulfil delivery and pickup orders, with little or no dine-in space and often no street-facing brand at all. Customers rarely set foot there; many never learn the address. The kitchen exists, for practical purposes, inside delivery apps and a packaging label.

The term covers a few distinct setups that are easy to confuse. Independent dark kitchens are a single operator cooking one or more menus from a low-cost site. Shared or commissary facilities rent out small commercial kitchen pods to multiple tenants under one roof, sometimes with shared cleaning, storage, and delivery-handoff logistics. And virtual brands are delivery-only menus that may be cooked inside an existing bricks-and-mortar restaurant, a burger joint quietly running a second wings concept from the same line.

The boom, and the shakeout

When dining rooms closed and delivery demand spiked, the delivery-only model looked like an obvious winner. Capital poured into shared-kitchen operators, and existing restaurants spun up virtual brands to capture extra app orders from idle kitchen hours. For a moment, the pitch was irresistible: all the demand, none of the dining-room overhead.

Then demand normalised as dining rooms reopened, and the structural problems surfaced. A lot of capacity had been built on the assumption that delivery would keep growing at pandemic rates. It didn't. Several high-profile shared-kitchen operators retrenched, closed sites, or exited markets entirely. Many hastily launched virtual brands quietly disappeared from the apps. The lesson was not that the model is broken, it's that it was massively oversupplied and often run on optimistic numbers.

The unit economics: where the money goes

The appeal is real on the cost side. A ghost kitchen skips prime retail rent, dining-room build-out, hosts, and servers. But the savings get eaten on the revenue side, and that is the whole story.

What you save

• Rent and fit-out: a back-street unit costs a fraction of a high-street storefront, and there's no dining room to furnish or maintain.
• Front-of-house labour: no waiters, hosts, or bar staff, kitchen and packing only.
• Flexibility: a virtual brand can launch, test, and be killed in weeks rather than signing a multi-year lease.

What you pay instead

• Delivery commissions: third-party marketplaces typically take a cut commonly cited in the 15-30% range per order, depending on the package and market. That comes straight off the top of every sale.
• Packaging and waste: delivery-only means every order is fully packaged, and food quality degrades in transit, which constrains the menu and adds cost.
• Discoverability: with no storefront and no foot traffic, you are entirely dependent on app placement, ratings, and paid promotion to be found at all.

Stack those against an industry where full-service restaurant net margins are often described as thin, frequently cited in the 3-9% range before everything goes right, and the problem is clear. Hand a meaningful slice of revenue to a delivery platform, layer on packaging, and pay to be discovered, and a delivery-only order can be less profitable than a dine-in cover, not more. For the deeper breakdown, see our guide to food delivery economics.

Saving on rent doesn't help if you hand the same money, or more, to a delivery app and a packaging supplier. The kitchen is cheap; the customer is expensive.

Where the delivery-only model still works

After the shakeout, the survivors share a pattern. They aren't betting the business on delivery-only economics from a standing start; they're using the model where it has a genuine structural edge.

1. 1.Spare kitchen capacity. An existing restaurant with idle line hours, a pizza kitchen quiet at lunch, say, can run a virtual brand at low marginal cost. The rent and core labour are already paid; incremental orders mostly cover food and commission. This is the most reliably profitable use of the idea.
2. 2.Established brands extending reach. A known restaurant can open a delivery-only satellite to cover a neighbourhood it can't reach from its main site, leaning on an existing reputation so discoverability isn't starting from zero.
3. 3.Dense urban catchments. Delivery economics improve with order density: short distances, high volume, and enough population within a tight radius to keep the kitchen busy and delivery times short.

What operators learned

The hard-won consensus is pragmatic. Brand still matters: a generic 'wings and fries' concept with no recognition has no reason to win an app search, and chasing volume with one more lookalike menu rarely pays. Platform dependence is the core risk, because the commission, the ranking, and the customer relationship all sit with the marketplace, not the kitchen. And the menu has to be engineered for transit, items that survive twenty minutes in a box, not delicate plates that arrive ruined.

The clearest verdict: delivery-only is a tactic, not a business model on its own. As an extension of something with real demand, it can be a smart margin play. As a way to launch a brand from nothing with no foot traffic and full commission exposure, it remains a difficult bet.

The bottom line

Ghost kitchens were never magic, and they aren't dead. The delivery-only model works best as a margin extension of demand that already exists, spare capacity, known brands, dense catchments, and works worst as a from-scratch bet that leans entirely on app placement and pays full freight on commission and packaging. Treat it as one tool among many, cost it honestly against the cut the platforms take, and the answer to 'does it still work?' is the same as for most things in this industry: yes, when the numbers do.