Every Startup Has a Landlord
Every startup has a landlord. Not an investor, who can decline to fund the next round but cannot stop you from operating. Not a customer, who can leave. The landlord is the party whose permission you are quietly renting, who can end your company with a letter, and who owes you no explanation when they do.
Most founders cannot name theirs. That is the problem.
What payments taught me
In payments the landlord was always the bank. Whatever the brochure said about your technology, nothing moved without a sponsoring bank willing to put its license behind your traffic, and the bank could withdraw that willingness for reasons that had nothing to do with you. A regulator asked a question. A board got nervous about a category. A new head of risk arrived and decided that fintech partnerships were more trouble than they were worth. The letter arrived, and a company that had spent two years building was given ninety days to find another bank, which is roughly the amount of time it takes to find out that every other bank has heard why you are looking.
I watched good companies spend a year on product and an afternoon on the bank agreement, and the afternoon decided more of their fate than the year did. The bank's risk appetite was their ceiling. The bank's compliance team was a stakeholder nobody had put on the org chart. And the product roadmap, however ambitious, was in practice a list of things the bank had not yet said no to.
The landlord changes, the lease does not
Every generation of startups has its own landlord, and each one believes, for a while, that it has escaped the pattern.
The mobile generation's landlord was the app store. A review policy was the law, an unexplained rejection was the eviction, and a change to the revenue split was a rent increase nobody could negotiate. The platform generation's landlord was the API that was open until the platform decided to build the feature itself. The cloud generation's landlord was quieter, because cloud providers rarely evict, but the bill is a lease, and anyone who has tried to move a mature system off one provider knows what the exit clause really says.
And now there is the model vendor. An AI company that is a thin, clever layer over one provider's model is a tenant in the oldest sense. The rate limit is the rent. The deprecation notice is the lease ending. The policy update that reclassifies your use case is the letter from the bank, arriving by email, with the same ninety days or fewer. I build on these models myself and I am grateful for them. I also know exactly what I am.
Rules for tenants
Name your landlord. Sit down and write a list of every party who could stop your company from operating by their own decision, without going to court. If the list is empty, you have missed one. Most companies have three or four.
Read the termination clause before the pricing page. Pricing tells you what success costs. Termination tells you how you die. The second is the one to memorize. How much notice? For what causes? What happens to your customers' data, your transactions in flight, your revenue in the settlement window, on the day it ends?
Your landlord's risk appetite is your ceiling. Before building anything, ask the landlord what they will never allow. Not what they allow today. What they are constitutionally unable to stomach. Build nothing that depends on that answer changing.
Have a second landlord before you need one. A second bank, a second model provider, a web version alongside the app. It costs more and it is slower, and founders are right that it is a tax on focus. It is also the difference between a renegotiation and a funeral. If you genuinely cannot have a second, know precisely why, and price that risk into every decision you make.
Never build your product on a feature the landlord also sells. The platform will ship it. The bank will offer it directly. The model vendor will add it to the product. If the only thing between you and your landlord is a feature they could build in a quarter, you are not a company. You are a product roadmap item they have not yet prioritized.
Treat the landlord's compliance team as a customer. They are the people who decide whether you exist. Learn their names. Answer their questions before they ask. Make their job easier than any other tenant does, because when the category gets reviewed, the tenant who was easy to supervise is the one who stays.
What this is not
This is not an argument against building on platforms. Nearly everything worth building is built on someone else's permission, and the alternative, owning your whole stack down to the license and the metal, is slower, more expensive, and often worse. The sponsor bank gave fintech a way to exist. The model vendors have given this generation a way to build things that would have taken a decade on their own. Renting is how most good things get started.
The argument is against forgetting that you rent. The founders I have watched get hurt were not naive about their technology or their market. They were naive about their lease. They had memorized the pitch deck and never read the clause that said how it ends, and when it ended, they discovered that the thing they had built was never entirely theirs.
The companies that survive are not the ones with no landlord. There is no such company. They are the ones who knew who held the lease, kept the relationship in good repair, built the second exit while the first was still open, and planned, from the first day, for the afternoon the terms changed.