By Craig Miles · June 2026 · 10 min read
The world has been building new schools in the developing world for more than sixty years. Since the 1960s, international development organisations, national governments, bilateral aid programmes, and philanthropic foundations have collectively invested hundreds of billions of dollars in school construction, teacher training, and education infrastructure. The results have been real; global primary enrolment has risen dramatically, and literacy rates have improved in almost every region.
And yet. In 2026, the number of children without access to education stands at 273 million, higher than it was in 2015, higher than it was in 2019, and higher than it was last year. For seven consecutive years, the out-of-school count has risen. UNESCO’s own modelling says that if current trends continue, universal upper secondary completion will not be achieved until 2105.
This is not a story about insufficient effort. Enormous effort has been expended. It is a story about a model that has reached the limits of what it can achieve — and a policy community that has not yet fully accepted that conclusion.
I want to make a precise argument here, because I think precision matters. I am not saying that schools are bad, or that teachers are unnecessary, or that physical education infrastructure has no value. Obviously, it does. I am saying something more specific: that the infrastructure model — build a school, hire a teacher, connect it to a curriculum, is structurally incapable of reaching the 273 million children who are currently outside its reach. And that continuing to invest primarily in that model, in the hope that it will eventually reach those children, is not a strategy. It is an article of faith that the evidence does not support.
Where the infrastructure model works, and where it breaks
The infrastructure model works extremely well under a specific set of conditions. It works when population density is high enough to make a school economically viable. It works when political stability is sufficient to keep a school open. It works when roads exist to get teachers and materials to the school. It works when the local economy can support a teaching workforce. It works when grid power is available to keep a school lit and functional.
Remove any one of those conditions and the model strains. Remove several, and it fails entirely.
Now look at where the 273 million out-of-school children actually are.
According to UNESCO’s 2026 GEM Report, the largest concentrations are in sub-Saharan Africa, where population growth is outpacing school construction, and Central and Southern Asia. A significant and undercounted proportion are in conflict-affected areas, where schools are destroyed, teachers flee, and the political conditions for building and maintaining educational infrastructure simply do not exist. UNESCO estimates that over one in six children globally live in conflict-affected areas, representing millions of additional out-of-school children beyond those captured in official statistics.
These are not populations that are almost within reach of the infrastructure model. They are populations that the infrastructure model is structurally incapable of serving, not because of funding shortfalls, but because the preconditions for the model to function are absent and, in many cases, cannot be created within any realistic timeframe.
The population growth problem
There is a second structural problem that receives less attention than it deserves: the infrastructure model cannot build fast enough.
Sub-Saharan Africa has the fastest-growing young population on Earth. The number of school-age children in the region is increasing year on year, and the rate of school construction, even with significant international investment, is not keeping pace. UNESCO’s data shows that since 2015, the out-of-school population in low-income countries has increased by 29%. This is not primarily because schools are being destroyed or closed. It is because the number of children who need schools is growing faster than the number of schools being built.
This is a race the infrastructure model cannot win. You cannot build your way out of a demographic trend of this scale, on this timeline, using the construction and staffing rates that are currently achievable.
That is not a counsel of despair. It is a statement of arithmetic. And arithmetic is useful, because it points clearly toward what kind of solution can actually work: one that scales faster than population growth, without requiring proportional increases in physical infrastructure.
The conflict problem
Conflict is the other structural barrier that the infrastructure model cannot solve, and it is growing, not shrinking.
UNESCO’s 2026 report specifically highlighted the Middle East, where ongoing regional tensions have forced schools to close and pushed millions of additional children out of education. But conflict-driven school exclusion is not a regional problem. It is a global one. Afghanistan, Sudan, the Democratic Republic of Congo, Myanmar, Mali, Somalia — in each of these countries, millions of children are out of school not because a school was never built, but because the school that existed has been closed, damaged, or made inaccessible by violence.
Building more schools in conflict zones does not solve this. A school is a fixed, visible, vulnerable asset. It can be closed by a government decree, damaged by an airstrike, or made inaccessible by a militia checkpoint. The infrastructure model, in conflict settings, is fragile in ways that are not addressable by better construction or more funding.
What is needed in conflict settings, as in remote rural settings, is an education delivery mechanism that does not depend on a fixed physical asset. One that cannot be closed by decree, cannot be damaged by an airstrike, and cannot be blocked by a checkpoint. One that exists in a child’s pocket and connects to the sky.
The teacher supply problem
Even setting aside conflict and geography, there is a third structural constraint: the world does not have enough teachers, and the shortfall is getting worse, not better.
UNESCO estimates that 44 million additional teachers will need to be recruited globally by 2030 to achieve universal primary and secondary education. That is not a shortfall that can be trained away in four years. Teacher training takes time, and retaining trained teachers in remote and low-income communities is consistently difficult — qualified teachers leave for urban areas where salaries, services, and quality of life are better.
This is not a criticism of teachers or of teacher training programmes. It is a description of a labour market reality that is not going to change on a timescale relevant to SDG4. The children who are out of school today will be adults before the teacher pipeline could conceivably close the gap through conventional means.
What the infrastructure model’s advocates get wrong
I want to be fair to the counterargument, because it is not entirely without merit.
The advocates of continued infrastructure investment point out — correctly — that physical schools provide more than curriculum delivery. They provide social infrastructure, peer learning, physical safety, nutrition programmes, and the social signalling that education matters in a community. These things are real. A child who attends a school learns things that a child with a device does not, and some of those things are not easily replicated by technology.
They also point out — again correctly — that technology-first approaches to education in the developing world have a poor track record. One Laptop Per Child, various e-learning platforms, tablet distribution programmes: many of these initiatives produced disappointing results, and the literature on why is extensive and sobering.
Both of these points are valid. But they do not support the conclusion that the infrastructure model should remain the primary vehicle for reaching the 273 million. They support a more nuanced conclusion: that technology-based education delivery must be designed differently from how it has been attempted before, with community integration, culturally appropriate content, adaptive pedagogy, and genuine follow-through. Not a device distribution programme. A system.
And crucially: the choice is not between a school and a device. For 273 million children, the choice is between a device and nothing. Because the school is not coming. Not by 2030. Possibly not by 2050. The infrastructure model, however much we invest in it, cannot reach them on any timescale that is relevant to their childhoods.
The 2105 number is the important one
Politicians and development organisations focus on 2030 — the SDG deadline. But the number that should be driving urgency is 2105: the year by which UNESCO projects that universal upper secondary completion will be achieved if current trends continue.
That is not a planning horizon. That is an indictment. It means that the children who are out of school today, and their children, and their grandchildren, and their great-grandchildren, will all live and die before the infrastructure model reaches everyone it needs to reach.
No serious person believes that is acceptable. But acceptance of the status quo is functionally what continued exclusive investment in the infrastructure model represents — because the infrastructure model is the status quo, and it is producing the trajectory that leads to 2105.
What changes that trajectory is not more of the same. It is a parallel delivery mechanism — not a replacement for schools where schools exist, but a genuine alternative for the populations that schools structurally cannot reach. One that scales with the satellite constellation overhead, not with the construction budget available. One that updates with the curriculum, not with the next school building programme. One that reaches a child today, not in eighty years.
What this means for how we think about investment
None of this is an argument against building schools. Where schools can be built, where teachers can be recruited and retained, where the political conditions are stable enough to sustain educational infrastructure — build the schools. The infrastructure model works well where its preconditions are met, and those preconditions exist in many places.
The argument is about the marginal investment — the next billion dollars of international development funding directed at education access. Where should it go? Into more school construction in the communities that are structurally hardest to reach via infrastructure? Or into the development and deployment of a satellite-connected, AI-tutored, device-based education system that can reach those communities today?
I think the answer is clear. And I think the development finance community is beginning to reach the same conclusion — not because they have abandoned their commitment to physical schools, but because the arithmetic of 2105 is becoming impossible to ignore.
The ReachED model is my attempt to describe what the alternative looks like in engineering terms — concrete, deployable, and built on technology that already works. If you want to understand the technical architecture in detail, the full explanation is in What Is the ReachED Model.
The buildings are not coming fast enough. The signal already is.