Build what others won’t
The queue for the obvious idea is long and expensive. The one nobody is standing in is neither.
Thesis
We are not looking for a category that is about to be hot. We are looking for a job that thousands of people already do badly, expensively, on paper, every single day — and that somebody is already paying for.
The underlying bet
Odds measure consensus, not potential. Every defining company, market and breakthrough started as a crazy bet some lunatic made.
Low odds don’t mean low outcomes — they mean low competition.
Why now
Six things we think are true at the same time. Together they are the reason we start companies now rather than wait for conditions to settle, because by the time they settle the opening has closed.
A shift, not a cycle
Steam, electricity, the internet. Every few generations the cost of doing something collapses and the old assumptions go with it. We are early in one of those.
Read the argument →Every industry is back in play
When the ground moves, incumbency stops being protection. What looks unassailable was built for conditions that no longer hold.
Read the argument →Building keeps getting cheaper
What needed thirty people and two years now needs a handful and a season. Anything priced on today’s cost of building is already priced wrong.
Read the argument →Small teams win this round
Scale used to be the advantage. Increasingly it is the tax. Five people who live with the problem outrun three hundred who must be briefed on it.
Read the argument →The opening is widest here
Emerging markets are not a smaller copy of the same opportunity. They are where almost nothing is built yet, and the defaults are still unset.
Read the argument →Behind, and closing the distance
Africa is behind at the frontier and the gap widens yearly. That is the hard part and the whole opportunity: the tools to close it finally exist here.
Read the argument →We go looking for
We pass on
Who we back
Everything else is negotiable. We have backed people with no company, no co-founder and no savings, and passed on polished teams with all three.
Convictions
The whole list. We disagree with parts of it roughly once a quarter and keep it anyway.
The queue for the obvious idea is long and expensive. The one nobody is standing in is neither.
By the time a category is hot, the interesting part of it has already been priced in.
If four other funds already agree with you, you are not early. You are on time, which is late.
It flows to the same cities, the same founder archetype, the same ideas everyone agrees are safe.
The pattern describes what already worked. It is a poor description of what will.
A ten-year fund life is an accounting decision. It is not how long a good company takes.
If it was never going to change anything, succeeding at it wasn’t worth the years either.
Every defining company started as a crazy bet some lunatic made against the numbers.
The question is never how big it is. It is what has to be true for it to become big.
A single good company in a thin market stays a single good company.
Low odds don’t mean low outcomes. They mean low competition.
Currently hunting
These are open briefs, not secrets. If one of them is the problem that has been annoying you for three years, we should talk.
The continent imports almost everything it uses and assembles almost none of it. Tooling, components and small-run machines get cheaper every year, and the companies that should be using them mostly do not exist yet. This is the least crowded thing on the list.
HardwareCompute needs power, and power is the binding constraint on everything downstream of it. Generation, storage and the data infrastructure sitting on top, built for markets that were never going to be served by somebody else's build-out.
InfrastructureClinics, plumbers, schools, workshops. Millions of small operators running on paper, or on six subscriptions they half use. One tool that does the whole job, filings included, so a team of five gets the output of fifty.
SoftwareCommerce and logistics end to end, with whatever the job needs: software, hardware, devices on the vehicle. First mile and last mile both cost too much, and in most of these businesses the cost of the mile is the business.
LogisticsEducation, science and research move at the speed of their tooling. Better instruments, better data and better software shorten the distance between a question and an answer, and that compounds harder than anything else here.
ScienceAfrica holds more human genetic variation than everywhere else combined, and almost none of it sits in the reference data drugs are designed against. A therapy validated on European genomes is a coin toss here. The sequencing, the reference panels, and the biotech research that starts from these genomes instead of correcting for them afterwards.
BiotechInformal trade, real prices, how goods move, what clinics see. Almost none of it lands in a dataset anyone can query, so every model built on the global corpora is confidently wrong about a billion people.
DataFilm, music, publishing, and everything that replaced them: creators, streaming, and the platforms underneath. The audience is already here and already paying attention. Almost none of the pipes that carry the work, or the rights that ride on them, are owned here.
MediaWhat we mean by “a wedge”
The smallest thing you can build in six months that somebody will pay for on its own merits — not because of the roadmap behind it. If explaining the value needs the second slide, it isn’t a wedge yet.
Rolling intake, no application form. A photo of the drawing is genuinely enough.