00Round one
Africa never had universal bank accounts.So it built mobile money instead.
It never had one national payment rail. So it built dozens: wallets, agent networks, USSD codes, each one solving access for people traditional banking had already written off.
That fragmentation wasn’t a footnote. It was the whole story.
Financial inclusion here didn’t happen because banks got better. It happened because someone built infrastructure for the reality on the ground (cash-based, mobile-first, allergic to hidden fees) instead of forcing that reality to bend toward a system designed in someone else’s country.
We’re watching the exact same pattern unfold in AI right now. Almost nobody building AI infrastructure has lived through round one closely enough to recognize it.
01Chapter one
Start with the token
A token from a given model is a weirdly clean unit of value. An OpenAI token behaves the same whether you hit it via OpenAI, Azure, or Bedrock. A Claude token behaves the same across its providers. Within a model family, it’s fungible, priced, tradable. Basically a currency. A store of value. A unit of account for what a person or an agent can actually get done.
This is already happening, quietly. Labs price in tokens. Developers budget in tokens. Agents will soon spend tokens the way people spend cash: continuously, autonomously, against a balance.
If tokens are becoming the currency of a new economy, the question is simple: who actually has access to that currency, and on what terms?
Because right now, AI access assumes you have
a bank accounta credit cardthe ability to transact in USD
That assumption was false for African financial services.
It’s false for AI access, for the exact same people.
A developer in Lagos or Nairobi who wants to build with AI has to route local-currency income through a card, through a bank, through cross-border FX, through a foreign biller, stacking cost and unpredictability at every hop before they’ve spent a single token.
- Local income
- Card
- Bank
- Cross-border FX
- Foreign biller
- Token
- Local currency
- router.africa
- Tokens
Meanwhile, the people actually driving commerce across the 18+ African markets we work in are transacting natively in mobile money, in local currency, almost entirely outside the card rails.
The infrastructure that made financial inclusion possible in Africa (local-currency rails, low-friction deposits, accessible pricing) doesn’t exist yet for AI.
That’s the gap router.africa is built to close.
02Chapter two
What we’re actually building
- 01
One API, one balance, across providers
- 02
Local currency in (shillings, naira, cedis, whatever you hold), tokens out
- 03
No foreign card, no international bank account required
This is infrastructure: the plumbing other products get built on top of. It makes money the way payments infrastructure always has:
Built on things we already understand deeply from Niobi, that don’t depend on anyone’s goodwill to survive.
Our first customers are developers, startups, and companies building with AI across African markets, the population infrastructure keeps forgetting to design for.
They’re the wedge.There’s a much bigger vision behind them.
03Chapter three
The bigger bet
Access to intelligence is on the same trajectory access to money was on a generation ago: scarce and gatekept, heading toward basic utility.
We’re already seeing a self-interested version of that: labs handing out free credits to juice adoption. That’s a growth tactic with an expiration date.
Real democratization looks like
in a village with no bank account
in San Francisco with a corporate card
Because the plumbing to deliver that access, cheaply and reliably, actually exists.
We don’t think you get to build that world by starting there.
You get there by laying rails that are commercially sound on their own first, serving developers and businesses who’ll pay for reliable, local-currency AI access, and using the distribution and trust that generates to eventually walk into a room with labs, governments, and development partners holding plumbing that already works, ready to carry a subsidized layer for students and underserved communities once it’s built.
- Step 01Rails that stand on their ownDevelopers and businesses who’ll pay for reliable, local-currency AI access.
- Step 02Distribution and trustEarned by doing the job, quietly, every day.
- Step 03A room with labs, governments, development partnersHolding plumbing that already works.
- Step 04A subsidized layerFor students and underserved communities, once it’s built.
04Chapter four
Why us, why now
We’re not trying to out-route or out-select the model layer. That game’s already being played well by other people, and it isn’t ours to win.
Our edge isn’t cleverness about which model is best. It’s that we know, structurally and operationally, how to move value into and out of fragmented African markets at scale.
The infrastructure speaks for itself, quietly doing its job, the same way mobile money never needed a marketing campaign to prove it worked. It just moved money for people who had nowhere else to move it.