Africa needs what Kai-Fu Lee's team did

One caveat before anything else, because the piece is not credible without it. When I say Africa I mostly mean Nigeria, because Nigeria is what I know. Fifty-four countries do not share a tech ecosystem and I am not going to pretend otherwise.

With that said. Nigerian tech, and a lot of African tech, is heavily concentrated in fintech and in building things that already exist in the West. There are not many genuinely new ideas being attempted. I do not say that as an insult. It is close to what China went through in the nineties, and in China copying turned out to be a training phase rather than an end state.

The difference is the bit that worries me. China had strong technical universities pushing talent out domestically, so the copying happened inside the country and the capability stayed there. African technical talent is still largely trained abroad. A brilliant kid from a village wins a scholarship, goes to Harvard, takes a job at Apple, and does not come back. Entirely rationally. So the answer has to be taking the training to the continent, teaching people to build and solve locally, at something like Andela's scale and considerably more of it.

The number that should frame all of this: the median age in Africa is around nineteen or twenty. In China and the United States it is around thirty-eight. That is the whole argument in one statistic. The largest young population on earth is arriving into a world where AI is ambient, and they already know what it is.

Now the part that changed my mind while I was working on this.

Every organisation I can think of that set out to train African technical talent has drifted away from training. Andela began by training and placing African developers and moved toward being a global talent marketplace. Moringa School in Kenya moved from bootcamp toward corporate and government partnerships. Decagon in Nigeria moved from pure training toward outsourced engineering services. Gebeya in Ethiopia went from training academy to talent marketplace. ALX has restructured repeatedly.

I am not claiming any of them did the wrong thing. Each of those moves is defensible on its own. What interests me is that they all moved in the same direction.

The common thread is that training alone does not pay. So everyone drifts toward selling the talent onward, or toward institutional contracts. And nobody, as far as I can see, has found a model where the value stays with the trainer, or with the local economy. The developer leaves for a Western salary, which is the correct decision for that developer and their family, and the capability leaves with them.

Which reframes the question. It is not should we train people in Africa, because nobody disagrees with that. It is: why does every attempt drift toward exporting talent, and what would have to be different for the value to stay?

I do not have the answer. I have three candidate models, and I would rather leave them open than pretend one of them is obviously right.

The first is an income share agreement. The trainee pays a percentage of salary for a fixed period after they are employed, so the value follows them abroad instead of being lost. It exists and it works domestically: Masai School in India runs free-until-employed then a salary percentage, and there are similar arrangements in Ukraine through outsourcing-linked academies. There are two serious caveats. In the United States, BloomTech ran this model and ended up under a CFPB consent order in 2024 for misleading students about what the agreements were, including telling them the agreement was not a loan when it carried an average finance charge of around four thousand dollars. The founder was banned from the student lending industry. It can work, but the terms have to be transparent and presented up front, and that failure mode belongs in any honest discussion of it. The second caveat is sharper and more specific to our case. Masai works partly because the graduate and the school sit in the same jurisdiction. Once the graduate is at Google in California and the school is in Lagos, collection becomes a cross-border legal problem that can eat the margin it exists to capture. Which is precisely the case the African version has to survive.

The second is equity or revenue participation in what graduates go on to build, which means funding company formation rather than employment. Harder, slower, and it only pays if some of them build something.

The third is sovereign or foundation-funded talent infrastructure, with no expected commercial return, because you treat talent the way you treat roads. There is precedent: China ran state-funded talent programmes deliberately, and the framing there was infrastructure rather than education. My honest caveat is that I do not see African education ministries funding this out of their existing budgets. But the framing is right and worth arguing for on its own.

Village Coder, the container schools I am building, is the third model. Free, philanthropic, no commercial return expected. Which is honest, and it leaves me with a question I have not resolved. If the only model I can personally make work is the one that requires somebody to give money away, does that mean handouts are the only thing that works?

I do not think that is the conclusion. But the fact that the question is still open is itself the finding, and I would rather publish it open than tie a bow on it.

“Nobody has found a model where the value stays with the trainer, or with the local economy.”