How to start a dark kitchen from scratch: the real cost vs. the quote you were handed

What does it cost to build a dark kitchen from scratch as of September 2026?

As of September 2026, building a dark kitchen from scratch costs between 8,000 and 45,000 USD, and that wide spread has nothing to do with kitchen size:

it depends on whether you rent already-equipped square meters or put up your own civil works. The 8,000 USD floor buys a station inside a shared hub, where the extraction hood, the grease trap and the health permit are already solved by the property operator, so you walk in with smallwares, a refrigerator and your virtual brand. The 45,000 ceiling shows up when you take a bare shell, because a hood with vertical ducting plus the fire-suppression system eat 6,000 to 11,000 USD before a single oven arrives. A client in Bogotá quoted 31,400 USD in March 2026, and 28,900 of it was steel, construction and extraction; the marketing line read «pending», and that empty line cost him the business.

What each investment range actually buys, line by line?

The three investment ranges buy different things, not different amounts of the same thing. Between 8,000 and 15,000 USD you get a station in a shared hub:

3,500 to 6,000 in used line equipment with warranty, 1,200 to 2,000 in smallwares and refrigeration, 900 to 1,500 for deposit and first month, and —this is what almost nobody budgets— 2,000 to 4,000 held back for paid media and menu photography. From 16,000 to 28,000 USD we are talking about a small own space with existing extraction, where you add 4,000 to 7,000 in electrical and gas work, plus a POS integrated with the apps at 600 to 1,100 a year. From 29,000 to 45,000 USD comes the full build: new hood, certified fire suppression, walk-in cooler and capacity for two or three virtual brands running on the same line.

What each investment range actually buys, line by line — in practice?

The global dark kitchen market grows 12.7% a year between 2025 and 2033, according to Global Growth Insights, and that growth makes good hub space pricier every quarter.

Five variables explain almost the whole gap between an 8,000 opening and a 45,000 one, and it pays to rank them by real weight. Extraction rules: if the property already has a certified hood and duct, you save 6,000 to 11,000 USD, a quarter of the entire project. Second comes the number of virtual brands, because each additional brand demands its own menu photography and health registration, roughly 1,100 to 2,200 USD apiece. Third, buying used instead of new cuts the steel line by 35% to 45%, with the caveat that a used combi oven resells at 55% of its value after two years. Fourth, the city and the delivery radius, which drive the lease deposit.

Five factors that move the price, and how much each one weighs

And fifth, the POS and aggregator integration: 600 to 1,100 USD a year, a minor line that decides whether you know your real food cost on Tuesday or on the 30th. A ghost kitchen's business is not in the kitchen; it is in showing up when someone hungry types «sushi near me» at 8:40 at night. That Bogotá project opened with a flawless line and billed 4,100 USD its first month against a break-even of 11,800: it was not short on kitchen, it was short on demand. Three dashboards decide that demand and hardly anyone budgets them. The ranking inside the delivery radius on Rappi and Uber Eats, which rewards whoever answers fast and punishes whoever keeps the courier waiting. The Google Business Profile listing —yes, a dark kitchen can and should have one if it takes pickup orders at a verifiable physical point—, because that is where «near me» searches land.

The mistake that kills the project is not in the kitchen

And geotargeted paid media, which buys your first 300 orders while the algorithm learns who you are. Set aside 2,500 to 5,000 USD for those three dashboards BEFORE you sign the lease. Between the wrong budget and the right one there is usually a difference of barely 7,400 USD, and yet the destination is worlds apart, because the method shifts money from a depreciating asset to an appreciating one. Steel loses value from the first service: that used combi oven sells at 55% after two years, and the walk-in is worth nothing outside the unit. Meanwhile, sales history inside the aggregator algorithm, accumulated reviews and a customer base with phone numbers gain value month after month. A listing with 340 reviews at 4.7 stars cannot be sold, yet it steadily lowers the acquisition cost of every new order, and 55% of restaurants report that their loyalty members' check grew faster than their menu prices, according to the Paytronix Loyalty Trends Report 2024.

Moving 7,400 USD between columns changes where the business ends up

Diego F. Parra hammers this point in every Masterestaurant diagnosis: buy steel last, buy demand first. With 32 dishes in a dark kitchen, average prep time at peak jumps to 27 minutes and you slide down the ranking, because both Rappi and Uber Eats penalize the restaurant that leaves the courier standing at the door. With 11 well-chosen dishes, that same time drops to the 12-to-15-minute range, inventory falls from 60 or 70 SKUs to under 30, and weekly waste stops eating two points of margin. Think about what happens if tomorrow you double the menu chasing coverage: every new dish adds a purchase SKU, a procedure and a stockout risk; at 20% stockouts the algorithm shows you less, less exposure means less volume, and less volume means you buy worse and your food cost climbs. The long menu gets paid for twice. Menu psychology techniques lift average check by 15% or more without touching prices, according to NeatMenu 2026, and they work better on short menus.

How to negotiate and optimize every line of the opening?

Negotiate the hub lease with a grace period and the equipment with a written warranty, in that order. Ask for two months free or rent stepped at 50% during ramp-up:

hubs sitting below 80% occupancy grant it, and that frees 1,400 to 2,600 USD for paid media exactly when you need it most. On equipment, buy used from closed restaurants with a 90-day written warranty and save 35% to 45%, except on refrigeration, where cheap turns expensive in electricity. Here I got it wrong for years: I used to recommend fighting app commissions before having volume, when that conversation belongs later, with some 1,200 monthly orders of history and a repeat rate that gives you an argument. Before that, the 27% commission is the rent on a commercial avenue you could not otherwise afford. This week, open your Google Business Profile and load 20 dish photos: it costs nothing and starts building the asset that actually appreciates.

Where the outcome is actually decided?

The investment gap between the two columns is only USD 7,400, yet the destinations could not be further apart, because the right method moves money out of a depreciating asset —steel— into an appreciating one:

sales history inside the algorithm, accumulated reviews and a customer base with phone numbers. A used combi oven resells at 55% of value after two years; a listing with 340 reviews at 4.7 stars never resells, but it lowers your cost per new order month after month. The long-menu mistake is the most expensive and the least obvious. With 32 dishes, average prep time at peak climbs to 27 minutes, and both Rappi and Uber Eats push down any restaurant that keeps couriers waiting; with 11 well-chosen dishes that time falls to 14 minutes, the app starts surfacing you higher, and the same kitchen sells more without another dollar of advertising.

Where the outcome is actually decided — in practice?

Shorter menu, more sales: counterintuitive, and the cheapest lever you own. One clarification on menus and QR codes, even if your business is 90% delivery:

the moment you allow pickup or set up a two-stool counter, keep the PHYSICAL menu alongside the QR one. The printed menu controls service rhythm, product storytelling and suggestive selling; the QR complements it by updating prices without reprinting, feeding delivery and giving you analytics on what people look at. Never QR only. Each has its role, and dropping the first one hands margin away for convenience. Diego F. Parra keeps hammering a point owners resist: in a dark kitchen, marketing is NOT a month-end expense, it is raw material, same as the chicken. No chicken, no dish; no order, same result. In the Masterestaurant model the digital budget enters the cost sheet as a fixed line from day one, with a target cost per new order that must stay under 12% of average ticket for the first 90 days.