Supply-Constrained
The most repeated piece of startup advice is “talk to your users.” For a lot of startups, it’s right. But the founders who built the most valuable companies of our generation didn’t discover what to build by talking to users. They already knew. The entire risk was whether they could build it.
Slack started as a gaming company called Tiny Speck. They built a multiplayer game called Glitch. It failed. But the internal chat tool they’d made to coordinate their team turned out to be something people wanted. They pivoted, iterated against user feedback, and built a $28 billion company. Every week in the early days was a new experiment. Does anyone want this? In what form? What do we charge?
SpaceX was nothing like that. Satellite operators were already spending billions a year on launch services. The demand was a line item in their budgets. Musk’s problem was building a rocket cheap enough to undercut the incumbents.
These are two fundamentally different types of ventures, and the startup ecosystem treats them as one.
Demand-constrained ventures face the risk that nobody wants the thing. Instagram pivoted from a location check-in app called Burbn after noticing users only cared about photo sharing. Airbnb had to convince strangers to sleep in other strangers’ homes. The founder’s job is searching for demand.
Supply-constrained ventures face only the risk of building it. Specific buyers are already spending specific money on specific transactions. Your product displaces those at better terms.
The test: can you identify, before founding, a buyer with a budget who is executing a transaction your product directly displaces? If yes, supply-constrained.
A natural objection is that precursor behavior exists for everything. People socialized before Facebook, traveled before Airbnb, communicated before Snapchat. But the distinction is more specific than behavior. When SpaceX launched, its customers could cancel an existing contract with Arianespace and sign one with SpaceX instead. The procurement process, the budget line, the vendor relationship all existed. Facebook had no equivalent. There was no purchase it intercepted. It created a new form of engagement and then discovered that advertisers would pay for the attention it generated. The product came first, the business model came later.
Look at the most valuable companies in history through this lens. Apple, Amazon, NVIDIA, Tesla. In every case, specific buyers were already executing specific transactions: buying phones and MP3 players, purchasing from catalogs, buying compute, buying cars. Each founder bet on building a dramatically better version of something people already paid for. The demand-constrained megasuccesses like Facebook are the exceptions. For every Facebook there are a thousand Clubhouses that searched for demand and never found it.
So where did the playbook come from? Y Combinator is the most influential startup institution in the world, and it is optimized for demand-constrained ventures. A $500K check tests whether anyone wants your SaaS product. It doesn’t build a satellite. Three-month batches require showing demand fast. Hundreds of companies per batch only works when each bet is cheap. The mentorship is calibrated to demand discovery: talk to users, iterate, measure retention. All of this works for what YC does. The problem is that YC’s cultural dominance turned this specific playbook into universal startup advice.
YC’s own fastest unicorn shows the limits. Starcloud hit $1.1 billion 17 months after completing the program. Data centers in space. Capital intensive, hardware-dependent, contingent on Starship economics. No customer discovery, no pivoting. AI companies are desperate for compute, terrestrial data centers are hitting energy limits, space has unlimited solar power and free cooling. Starcloud became YC’s fastest unicorn while violating nearly every principle the playbook teaches.
The founder profile follows from the problem type. Demand-constrained founders solve a behavior puzzle: what do people want, how will they act, what will they pay for. Supply-constrained founders solve an engineering puzzle: which architecture, which path, can this be built at all. They need to be deeply technical and make the right bets on problems with no guaranteed solution. They still need to sell, but they’re selling vision to investors. Musk nearly lost SpaceX because he ran out of cash after three failed launches. The engineering worked on the fourth attempt. Convincing capital to fund something expensive with long timelines before revenue exists is an enormous sales job.
The startup ecosystem built its entire advice infrastructure around demand discovery. The companies that created the most value never needed to discover demand. They needed to build.