Why Ghost Kitchens Fail: 7 Breaking Points to Watch For

Ghost kitchens fail for reasons that show up in the P&L long before the doors close. The short answer: a delivery-only kitchen pays a marketplace commission of 15% to 30% on every single order it sells, and a food business running on a 10% margin cannot absorb that unless something else in the structure is unusually strong. Almost every other failure cause in this article is a variation on that one.

The collapse was not quiet. Funding for the category dropped 95% year over year in Q4 2025, from $210 million to $10.5 million, according to QSR Pro’s review of what survived. CloudKitchens shut its East Oakland facility in November 2025 and delayed its planned Middle East IPO. Local Kitchens closed more than half its locations in October 2025. Earlier, Uber Eats removed 8,000 virtual brands from its platform between March and July 2023, and Kitchen United closed all eight of its Kroger in-store locations in November 2023, per GrowthFactor’s category analysis.

None of that means the model never works. It means the model works under specific conditions, and this post is about what happens when those conditions are missing. For the other half of the picture, our breakdown of profitable ghost restaurants walks the unit economics of the kitchens that do clear a margin, and which levers get them there.

What this guide adds: Most posts on this topic list generic small-business failure causes and attach them to ghost kitchens. This one works from documented closures and the specific structural problems industry analysts identified afterward, then translates each into a signal you can check in your own numbers this month.

The Bottom Line

• The core failure is arithmetic: a 25% to 30% commission applied to every order in a business with a 10% margin does not leave a business.

• Ghost kitchens stack costs that a normal restaurant does not carry at once: platform commission, a marked-up sublease rate, and often percentage-of-sales rent on the same order.

• Food that degrades in transit loses the customer permanently, and a delivery-only kitchen has no dining room to win them back in.

• Operators who never build a direct ordering channel remain fully exposed to platform decisions they do not control.

• The delivery-only kitchens still running in 2026 almost all had an existing asset first: a brand, a kitchen already paid for, or several concepts sharing one line.

How often do ghost kitchens fail, and what does failure look like?

There is no clean national closure statistic for this category, partly because most delivery-only brands are launched inside somebody else’s facility and simply stop appearing on the apps rather than filing anything.

What is documented is the contraction of the facility operators that the whole model depended on. GrowthFactor’s timeline reads as one continuous retreat:

• January 2022: Reef Technology paused operations at 95 underperforming ghost kitchens, roughly a third of its footprint.

• May 2022: Reef cut approximately 750 jobs.

• Through Q2 2023: Wendy’s cut its Reef-based target from 700 planned locations to 100-150, then closed the remaining 15 or so US units.

• March to July 2023: Uber Eats removed 8,000 virtual brands from the platform.

• September 2023: CloudKitchens was reported at 50% occupancy across its buildings.

• November 2023: Kitchen United closed all eight Kroger in-store locations, later selling its intellectual property.

Alongside that, IBISWorld put the US ghost kitchen market at $2.9 billion in 2025 with revenue down 5.2% in 2024. This is a category that shrank while food delivery overall grew, which tells you the problem is internal to the model rather than a demand problem.

1. Delivery commission does not survive a ghost kitchen’s thin margin

A dine-in restaurant pays marketplace commission on the portion of its sales that arrive through an app. A delivery-only kitchen pays it on everything.

Published 2026 rates run about 15% for self-delivery plans, 25% for partner delivery, and 30% for premium placement across DoorDash, Uber Eats, and Grubhub. OPA’s analysis puts the real all-in cost at 30% to 40% per order once processing and promotion fees are included, reaching 43% in high-regulation markets like New York City.

Labrador AI’s margin breakdown shows where that lands: at 25% commission a well-run kitchen holds 5% to 7% on a delivery order, and at 30% it holds 0% to 2%. Run a whole business on 0% to 2% contribution and any bad month is terminal.

The signal to watch: calculate your blended commission rate across all orders for last month, not your plan’s advertised rate. If it is above 25% and more than 80% of your orders come through marketplaces, the business is running on volume rather than profit.

2. Cost layering stacks three claims onto the same delivery order

This is the failure cause that surprised operators most, and Restaurant Dive’s teardown of the category names it directly as cost layering.

A single order in a hosted ghost kitchen can carry all of the following at once:

• The delivery platform’s commission.

• A sublease rate priced above standard commercial per-square-foot, because the facility is selling convenience and speed to open.

• Percentage-of-sales rent on top of that base rate, so growing revenue increases the landlord’s take.

• A menu price marked up to cover the commission, which makes the food less competitive against a nearby restaurant that has not marked anything up.

Each layer is defensible on its own. Together they are the reason a kitchen can be a facility’s best-performing tenant and still not clear a profit, which is exactly what one operator quoted by QSR Pro described before closing.

The signal to watch: add your facility’s percentage rent to your blended commission rate. That combined number is what comes off the top of every order before you have bought a single ingredient.

3. Food quality falls apart in transit, and the customer does not come back

Plenty of dishes simply do not survive 25 minutes in a bag. Restaurant Dive’s reporting cites pho arriving cold with the ingredients separated as the type of failure that ends a customer relationship, noting it is “very, very hard to get them to come back” once that happens.

A dine-in restaurant that serves one bad plate can fix it at the table. A delivery-only kitchen gets one attempt, through a courier it did not choose, with no chance to intervene. And because the apps surface ratings prominently, a run of quality complaints suppresses the visibility the kitchen depends on for orders.

The operators who get this right narrow the menu to items that travel: fried food that stays crisp, bowls and wraps that hold structure, anything that is as good at 20 minutes as at 2. Steaks, delicate salads, and broth dishes are where this breaks.

The signal to watch: your marketplace rating trend by item, not overall. One dish dragging down a 4.6 average is usually a travel problem, not a cooking problem, and removing it is cheaper than defending it.

4. Platform dependency means the kitchen never owns its customers

When Uber Eats delisted 8,000 virtual brands in 2023, those brands did not get a vote. They also, in most cases, had no way to tell their customers where to find them, because the customer relationship belonged to the app.

This is the structural weakness behind most of the sudden closures. An operator with no direct channel has no email list, no order history, no repeat-purchase mechanism, and no bargaining power in a commission negotiation. Everything about the business, including whether it appears in search results at all, sits in somebody else’s product decisions.

The signal to watch: what share of your orders arrived through a channel you own, this month? If the answer is zero, you do not have a customer base, you have a supplier relationship with a marketplace.

5. A delivery-only kitchen has no free way to be discovered

A restaurant with a storefront gets discovered by people walking past, every day, at no cost. That awareness is worth real money and nobody puts it on a balance sheet until it is gone.

A ghost kitchen has no sign anybody sees. Discovery has to be purchased, either as a higher commission tier for better placement in the app or as advertising spend. Restaurant Business Online’s reporting on small operators trying the model found marketing difficulty and delivery cost cited repeatedly as the reasons it did not work for them.

This is why the same delivery-only concept succeeds for an established chain and fails for an independent. Chipotle and Sweetgreen open delivery-only sites carrying brand recognition and marketing budget they already paid for. An unknown brand launching from a rented kitchen is buying awareness at retail from day one, while already down 25% on every order.

The signal to watch: your fully loaded customer acquisition cost, including the commission premium you pay for placement. Compare it against the lifetime value of a delivery customer who has ordered from you twice. For most independents the first number is larger.

6. Running several virtual brands from one kitchen line multiplies mistakes

The multi-brand idea is sound in theory: spread one kitchen’s fixed cost across four concepts and the rent per revenue dollar drops.

In practice, Restaurant Dive found that multiple brands with different preparation requirements coming off the same line produced “mistakes and inconsistency.” Staff preparing tacos, pizza, and sandwiches at once made execution errors, and workers in shared facilities often had no familiarity with any individual brand’s standards because they had not been trained in that brand’s environment.

The kitchens where this does work run brands that share ingredients and technique, hold each menu to a tight item count, and staff a line trained on the whole portfolio. Four unrelated cuisines on one griddle is a quality problem dressed up as an efficiency gain.

The signal to watch: order accuracy and refund rate per brand. If one concept is generating most of the remakes, the line is overloaded rather than the recipe being wrong.

7. Ghost kitchen volume forecasts were built on a pandemic spike

Restaurant Dive describes pandemic-era delivery demand as a “false positive.” Dining rooms were closed, delivery was the only option, and the resulting order volume looked like a permanent behavior change. Facility operators and brands expanded against that curve, then the curve normalized.

The expansion was aggressive enough that its unwinding is the story of the category: Reef’s 95 paused kitchens, Wendy’s cut from 700 planned units to a handful, Wonder pivoting toward a more conventional model. Market power has since consolidated into a few groups, and the hybrid facilities that survived added pickup counters and seating, which is a quiet acknowledgment that pure delivery-only was not enough.

The signal to watch: whether your own volume forecast is built on your actual trailing 12 weeks or on a category growth rate from a market research headline. Global forecasts project this category past $177 billion. The US market, per IBISWorld, is $2.9 billion and recently shrank. Only one of those numbers has anything to do with how many orders your kitchen will get.

What the ghost kitchens that survived did differently

Three groups are still operating at scale, per QSR Pro’s 2026 review, and the pattern across them is consistent:

• Established brands using delivery-only sites as an expansion tactic. They bring existing recognition and marketing, so discovery is not a cost they have to solve.

• Multi-brand operators, particularly in Asia-Pacific, which captured roughly 48% of global ghost kitchen revenue by owning their brands outright and spreading fixed cost across concepts.

• Hybrid facilities with a pickup or seating component, which reduces the share of orders exposed to commission.

The unifying lesson is that a delivery-only kitchen works as a tool for a food business that already has something, and rarely as the business itself. A bakery, caterer, or restaurant with idle kitchen hours and a customer list can add a delivery-only brand cheaply, because the two expensive problems, occupancy and discovery, are already solved.

The other common trait among survivors: they built a direct ordering channel and handled at least some deliveries themselves. That is the only move that changes the commission arithmetic rather than working around it. Once a customer orders directly, the delivery becomes a logistics question instead of a 25% fee, and platforms built for local food businesses, Metrobi among them, exist to cover that side with multi-stop route planning and drivers who can work with the same business over time.

Are ghost kitchens still worth it in 2026?

For an independent operator with no brand, no kitchen, and no customer list, the evidence says no, and it says so clearly. The economics documented above do not resolve with better execution, because the commission applies to every order regardless of how well you cook.

For a food business that already owns a kitchen with slow hours, a name customers know, or a way to get its own orders delivered, the answer is different. That operator is not betting the business on delivery-only. They are adding revenue on top of fixed costs already being paid, which is the version of this model that has consistently worked.

Understanding why ghost kitchens fail is mostly about being honest regarding which of those two you are. The first group is buying revenue at a loss and calling it growth. The second is using an efficient tool. The kitchen equipment looks identical either way, which is what made this so easy to get wrong.