Build vs rent: the real cost of a central production unit in London

Most growing food businesses treat the central production unit as a property decision: find the space, price the fit-out, sign the lease. Framed that way, it sounds like a question about kitchens: build or rent?

It isn't. It's a question about capital.

The moment you commit six or seven figures to building your own production kitchen, that money stops being available for the things that actually grow the business: new sites, new channels, marketing, people. It converts from working capital into a fixed, illiquid, single-purpose asset. And in a business scaling, capital that's tied up in bricks is capital that's stopped working.

So the real question isn't "should we build or rent?" It's "where should our money be working over the next five years: in a building, or in the business?" We’ll walk through the true cost of each, with the London specifics that change the maths.

First, what is a CPU?

A central production unit, a commissary in the language most serious operators use, is a dedicated space for batch prep, cooking, chilling, packing and dispatch that feeds multiple sites, channels or contracts from one hub. It's how a multi-site restaurant group holds consistency across locations, how a corporate caterer produces at volume, and how a food manufacturer reaches audit-ready scale.

It is not a delivery-only "dark kitchen," and the distinction matters for this decision: a CPU is a production asset serving an operation, not a rented address for fulfilling orders. When we talk about build vs rent, we mean the real infrastructure (extraction, power, cold chain, drainage) that production at scale demands.

That infrastructure is exactly why building is expensive.

The real cost of building your own

The headline most operators anchor on is rent per square foot. It's the least important number in the decision.

On our 2026 cost modelling, building a production facility from a bare shell in London runs to roughly £273 per sq ft in upfront capital once you add professional fees, planning, power and the fit-out itself. For a mid-sized CPU of 2,000–3,000 sq ft, that's £550,000–£820,000 of capital committed before you produce a single unit. And that figure assumes a clean run: it's modelled on a good-quality modern unit with utilities already in place and no build complications, so a lower-grade building, a required substation upgrade, or a planning refusal pushes the real number higher.

What sits inside that number, and why London makes each line worse:

• Design and professional fees: architecture, M&E design, food-safety consultancy, and the planning and building-control submissions that vary borough by borough.
• Base build and finishes: drainage, hygienic flooring and wall cladding, sealed surfaces to audit standard.
• Extraction and ventilation: the most variable line, and the one boroughs scrutinise hardest near residential areas.
• Power: three-phase upgrades that depend on the DNO's timeline, not yours, which is why power is so often the item that stalls a build.
• Specialist equipment: combi-ovens, bratt pans, blast chillers, walk-in cold rooms and prep lines.

Two things make London specifically worse than the raw figure suggests. Extraction and power capacity are scarce and contested, so the two most expensive line items are also the two most likely to blow your timeline. And planning is borough-by-borough: what's waved through in one is refused in the next, and an extraction or change-of-use refusal can cost months and a full redesign.

But the fit-out cost isn't the real cost. Two larger costs sit underneath it.

The costs that never make the headline comparison

Time is the first. A ground-up London build realistically takes 13 months or more from heads of terms to first production batch, and complex sites stretch well beyond that. Throughout that period you're paying twice: the carry cost of the new site (rent, rates, financing on committed capital) and the cost of the old setup you haven't yet left. That overlap, not the fit-out invoice, is what quietly drains cash during a build, and it's invisible in any rent-per-square-foot comparison. Contrast that with moving into space that's already licensed and audit-ready, where production can start in about two weeks.

The end of the lease is the second. Dilapidations on a heavily fitted kitchen (stripping out extraction, reinstating finishes, returning the shell to its original state) are a significant liability that lands years after the build, and one most operators never model at all.

And the running costs are higher than the rent line implies. Add up the true annual cost of occupying and running a self-built London facility and it comes to around £110 per sq ft per year, and that's everything in: the base lease on the warehouse, business rates, water, waste and grease management, extraction and equipment servicing, pest control, building insurance, and security. The headline rent is barely half of it; the rest is the stack of separate contracts a self-managed unit quietly accumulates. (Again, this is a best-case estimate on a good-quality unit, so the real figure tends to climb, not fall.)

None of this argues that building is irrational. It argues that the headline comparison is the wrong comparison, and that the honest number is far larger, and far riskier, than most operators model.

What renting changes, and what it doesn’t

Start with what renting won't do, because this is where most comparisons oversell.

Renting won't transform your running costs. At the size we're discussing, a fully serviced unit at Karma comes to about £8.84 per sq ft per month all-in, against roughly £9.19 to occupy and run the equivalent facility you built yourself. Cheaper, yes, but by four per cent, and on modelling that already gives the build the benefit of every doubt. 

Look at how those two nearly identical numbers are assembled, though, and something more interesting appears.

Our rent is 64% higher than the self-build's. Our services are three and a half times cheaper. The self-build wins decisively on the number you negotiate, and loses all of it back on the numbers you don't, because services bought once for a single unit cost far more than the same services bought across six sites. That's not a pricing trick, it's just scale: one operator buying one pest contract will never beat six sites buying together and splitting it across every kitchen in those sites.

So if the monthly isn't the argument, what is?

It's roughly the same running cost without the £682,500 upfront and the thirteen months. That's the entire proposition (alongside a four percent on the monthly).

Which means what renting actually changes is everything sitting around that near-identical number:

• No capital. Nothing to finance, nothing depreciating, nothing locked in a building.
• No wait. Producing in about two weeks rather than 13-plus months, so the old, more expensive setup switches off far sooner.
• No dilapidations. No six-figure bill arriving years later to hand the shell back.
• No management load. Seven service contracts, seven renewal cycles and seven suppliers become one invoice and someone else's problem.
• Optionality. A ten-year lease on a bespoke build is a fixed bet on a single forecast. A flexible unit lets you scale up for a new contract, add space as you grow, or change direction if the business does. A building you own will do none of that.

This is the model Karma Kitchen is built on: fully fitted central production units across six London sites, with the infrastructure, servicing and compliance carried across all of them rather than by you alone.

Two things this piece doesn't cover in depth: whether a shared facility can meet BRCGS/SALSA and multi-site audit standards, and how production infrastructure affects your ESG and Scope 3 position. Both are real parts of the decision, so see how we support compliance for food manufacturers → and our approach to sustainability and Scope 3 →.

A worked comparison

Take a 2,500 sq ft CPU producing at volume for a multi-site operator, modelled over five years at Karma's standard rate. The figures below use our 2026 cost modelling; treat them as indicative, you can run your own numbers with our ROI Calculator.

Over five years the difference is roughly £735,000, and 93% of it is the capital you never spent. The running-cost saving accounts for the other seven.

That's the whole point, and it's worth sitting with. Building doesn't buy you a cheaper kitchen. It buys you the same kitchen, thirteen months later, with £682,500 of your capital locked inside it and a dilapidations bill waiting at the end. The question was never whether you could afford the build. It's what that £682,500 would have done if you'd spent it on the business instead.

Want to model your own size and lease term? Run the build-vs-rent calculator 

When building still makes sense

Honesty matters here, because the answer isn't always "rent." Building your own can be the right call when:

• you're a large, well-capitalised operator needing 20,000+ sq ft of fully bespoke process plant, where the economics genuinely tip towards ownership;
• your production is so specialised (high-care, allergen-free, particular manufacturing processes) that it can't sit within a multi-tenant environment; or
• you have a genuine 15–20 year, single-site horizon where seven-figure capex amortises over decades and long-term control is worth more than flexibility.

Even then, many operators rent first to prove volumes and de-risk the model before committing to a purpose-built factory, because the most expensive way to discover your forecast was wrong is to have already built for it.

For most mid-market caterers, restaurant groups below roughly 30 sites, and scaling food brands, staying asset-light for longer is simply the stronger financial position.

A decision checklist

Before you commit either way, work through four questions:

1. Capital. Can you lock £550,000–£820,000 into kitchen infrastructure without constraining site roll-outs or working capital, and does that beat what the same capital would earn deployed into the business?
2. Time. Can your current operation carry the cost of running two setups for 13-plus months if the build runs late (and builds run late)?
3. Risk. How certain is your production volume for the next 3–5 years, and what's your exposure if extraction is refused or power is delayed?
4. Flexibility. How likely are you to need more space, a different location, or a changed model within 2–3 years, and would keeping capital liquid preserve options you'll want?

If the honest answers point to uncertain volumes, tight capital, or a business still changing shape, the decision has usually made itself.

The bottom line

Whether centralising production makes sense is rarely the real question: for a scaling multi-site operator, caterer or manufacturer, it almost always does. The real question is how you access it. By tying up the best part of £700,000 and a year of runway in a building you'll eventually pay to dismantle, or by keeping that capital working in the business and producing within weeks, for much the same monthly cost.

Put plainly: a kitchen you build isn't an asset. It's capital that's stopped working. The operators pulling ahead are the ones who worked that out.

Thinking through a production move? Talk to us about your volumes and tour the relevant sites →. Most operators walk a space and have a clear answer within a fortnight.