Twenty years ago, if you were building a software company, you built a server room. You bought the machines, cooled them, powered them, staffed them, and when you grew you bought more. Owning your infrastructure wasn't a strategy. It was just the cost of being in business.
Then Amazon asked a question that sounded almost stupid at the time: why does every company build its own? The computing power was the same for everyone. The expertise to run it well was scarce and expensive. The capital tied up in it did nothing a customer could see. So Amazon built the infrastructure once, at scale, and rented it out by the hour. We call it the cloud now, and it's hard to remember it was ever controversial. But it was. Serious engineers argued you couldn't possibly run a real business on someone else's machines.
They were wrong, and the reason they were wrong is the same reason this matters for food.
The pattern is older than the cloud
The cloud wasn't the first time an industry unbundled its infrastructure, and it won't be the last. It's a pattern that repeats whenever three things are true.
Logistics went through it. Companies used to own their warehouses and run their own fleets, because how else would your goods move? Then third-party logistics providers, the 3PLs, built the warehousing and distribution once, at scale, and let everyone else rent it. Today a brand can reach the whole country without owning a single lorry or a square foot of warehouse.
The pattern is always the same. An industry treats owning its infrastructure as simply the way things are done. Then someone notices that the infrastructure is expensive, hard to run well, and identical for everyone who needs it, and that the real business, the thing customers actually value, is something else entirely. So the infrastructure gets built once, by specialists, and rented by the people who used to build their own. And within a decade, owning it looks less like a strength and more like a burden you took on before you knew better.
Three conditions make it happen: the infrastructure is capital-heavy, it's genuinely hard to run to a high standard, and it's not the thing your customers are buying. When all three are true, unbundling isn't just possible. It becomes inevitable, because the businesses that stop carrying the burden move faster than the ones that don't.
Food production has all three.
Why the kitchen is next
Think about what a production kitchen actually is to a growing food business.
It's capital-heavy: a fitted commercial kitchen runs to hundreds of pounds per square foot before you produce anything, and the serious ones run past seven figures. It's genuinely hard to run to standard: the extraction, the power, the drainage, the compliance, the audits, the maintenance, none of it is your team's core skill, and getting it wrong is slow and expensive to fix. And, crucially, it is not the thing your customers are buying. Nobody chooses a restaurant because of the drainage falls in its production kitchen, or a food brand because of how many phases of power its facility draws.
What customers buy is the product and the brand: the bun, the broth, the curry paste, the thing folded in front of them, the name they trust. The kitchen behind it is infrastructure. Necessary, expensive, invisible infrastructure, exactly the kind that every other industry eventually stopped insisting on owning.
So the same question Amazon asked about servers is now worth asking about production: why does every food business build its own?
For most of the history of the industry, there was no alternative, so the question didn't mean anything. You needed to produce, so you built somewhere to produce, the same way a software company in 2003 needed computing, so it built a server room. The alternative simply didn't exist yet. What's changed is that it now does. The infrastructure can be built once, at scale, by specialists, and rented by the businesses that used to build their own.
What unbundling actually changes
The objection, when the cloud arrived, was that owning your own servers gave you control. And it did. It also gave you a data centre to run when what you actually wanted to do was write software.
The same is true here, and it's worth being honest about the trade. Owning your own production facility gives you total control over every detail. It also gives you a building to build, finance, staff, maintain, audit and eventually hand back, when what you actually wanted to do was make and sell food. Control has a cost, and for most of the history of the industry that cost was unavoidable. It isn't any more.
What unbundling changes is where a food business's capital, attention and risk go. Instead of sinking six or seven figures into a building and the months it takes to make it work, the capital stays in the business, funding the product, the brand and the next site, the things customers actually value. Instead of your best people managing extraction contracts and audit findings, they're doing the work you're known for. Instead of a decade-long bet on a single building sized to today's forecast, you scale the infrastructure up and down as the business moves.
None of that is a discount. It's a reallocation, of money, of attention, of risk, away from the infrastructure and towards the business. Which is exactly what the cloud did for software, and 3PLs did for retail.
The businesses that see it first
In every one of these shifts, some businesses saw the pattern early and some argued against it until it was obvious. The early movers weren't the biggest or the best-funded. They were the ones who worked out soonest that owning the infrastructure was never the point, that the point was the thing the infrastructure produced, and that carrying the building was slowing them down.
We're at that stage now with food. The operators who are scaling fastest are increasingly the ones who've stopped treating a production facility as something they must own, and started treating it as something they can rent, the same way they already rent their computing, their logistics, and half the other infrastructure their business runs on.
The kitchen is just the next thing to unbundle. The cloud came for computing, 3PLs came for logistics, and the same logic, capital-heavy, hard to run, not what the customer's buying, has arrived at the production kitchen. The question isn't whether food production follows the pattern. It's who works it out first.
This is the thinking behind Karma Kitchen: production infrastructure, built once and run well, so food businesses can rent it instead of building their own. Talk to us about your production →.
