Five Thousand Dollars
14 January 2026 - Hugo O’Connor
The entire venture capital model was built on a specific assumption about the cost of building software. You needed engineers, and engineers were expensive. You needed infrastructure, and infrastructure was expensive. You needed time, and time—measured in months or years of runway—was the most expensive thing of all.
So the math worked: raise $5 million, hire a team, build for 18 months, hope you find product-market fit before the money runs out. The VC model made sense because the capital requirements matched the risk. Big bets on ambitious problems that required serious resources.
But what happens when the cost structure collapses?
I’ve been watching this happen in real time. Projects that would have required a team of engineers and months of development can now be built by one person in weeks. Sometimes days. The LLM handles the code. The LLM handles the tests. The LLM handles the documentation. Infrastructure costs have already fallen to near-zero for most applications. What’s left?
Just the cost of figuring out what to build and whether anyone wants it.
This creates space for something that hasn’t really existed before: micro-VC. Not $5 million seed rounds. $5,000 seed rounds.
Five thousand dollars is enough to:
- Pay for API costs while you build and iterate
- Cover basic infrastructure for a year
- Give you enough runway to validate an idea without other obligations
- Buy focus for a month or two
But here’s what’s interesting: you might not need those things anymore.
The traditional VC model optimizes for scale. You raise big rounds because you need to move fast, capture market share before competitors, build moats through network effects or capital intensity. The entire game is predicated on winner-take-all dynamics in large markets.
But most problems aren’t like that. Most problems are specific, tractable, and serve defined communities. Most businesses don’t need to scale to unicorn status to be valuable. They just need to work and generate revenue.
With LLMs reducing the cost of building to near-zero, the bottleneck isn’t implementation anymore. It’s learning. Can you figure out what people actually need? Can you iterate quickly enough to find product-market fit? Can you build something people will pay for before your small amount of capital runs out?
These are different constraints than the traditional startup game. They favor different skills: deep domain knowledge, tight customer feedback loops, rapid iteration, and the discipline to keep things simple. You’re not building for massive scale. You’re building to solve a real problem for real people and get to cash-flow positive quickly.
The economics change everything. If you can build something viable for $5k instead of $5m, you don’t need to return 100x to make the investment worthwhile. A successful exit could be $500k instead of $500m. The founder might keep 90% instead of 10%. The business might generate steady cash flow instead of pursuing hypergrowth.
This isn’t traditional venture capital at all. It’s something else—closer to patronage, or grants, or creative funding. You’re not buying equity in the next unicorn. You’re enabling someone to spend a month or two exploring an idea that might not exist otherwise.
Some of these ideas will fail. Most of them, probably. But the failure cost is minimal. $5k is less than many companies spend on a single conference sponsorship. It’s a rounding error in a traditional VC fund’s operating budget.
And some of them will work. They’ll find paying customers quickly. They’ll iterate into cash-flow positive businesses that serve real needs. They won’t make anyone fabulously wealthy, but they’ll sustain themselves and grow organically. They’ll prove that there’s a vast space of valuable software that traditional VC has ignored because it didn’t fit the model.
I think there’s a broader shift happening here. We’re moving from an era where building software required enormous capital to an era where it requires almost none. The constraint isn’t resources anymore—it’s direction, taste, and the ability to learn quickly from real users.
- Solo founders become viable again (or viable at scale for the first time)
- The time to cash-flow positive shrinks from years to months
- Geographic and network advantages matter less
- Domain expertise matters more
- The risk/reward ratio for experimentation changes completely
Traditional VCs won’t chase these opportunities. The fund economics don’t work. A $100m fund can’t write $5k checks—the administrative overhead alone would kill them. But someone else could. A new model could emerge that’s specifically designed for this: micro-funds deploying small amounts of capital to large numbers of experiments, betting on a portfolio approach where a 10% success rate would be extraordinary.
Or maybe no formal funding is needed at all. Maybe $5k is small enough that individuals just fund themselves, or friends fund friends, or small communities pool resources to back people working on problems they care about. Maybe the entire intermediary layer disappears.
The traditional startup ecosystem—accelerators, pitch competitions, angel rounds, Series A through Z—was built for a world where software was expensive to make. That world is ending. We don’t yet know what replaces it, but the shape is starting to emerge.
It looks smaller. More distributed. More accessible. Less reliant on institutional gatekeepers.