Introducing: State of AI Diffusion
A new quarterly newsletter tracking how artificial intelligence is spreading across the African continent and low-to-middle-income countries — who has access, who doesn’t, and what we are doing about it.
Welcome to State of AI Diffusion, a quarterly briefing from the Equiano Institute on how artificial intelligence is spreading across the African continent and low-to-middle-income countries more broadly.
This is not a safety report. We will not map alignment failures or frontier model risks here — those matter, and they are covered elsewhere. We are asking a prior question: Who has access to AI’s benefits? Who doesn’t? And what are we doing about it?
Why now? Why this gap?
The global AI discourse is crowded. Reports on AI safety abound — from OpenAI, Anthropic, DeepSeek, and major research institutions. There is a flood of analysis on misuse, autonomy, and superintelligence. Those conversations are necessary. They miss something foundational:
Most of the world doesn’t have access to advanced AI yet, and the gap is widening.
According to Microsoft’s latest AI Diffusion Index, no African country had reached 20% AI adoption by the end of 2025. South Africa leads the continent at 21.1%, followed by Egypt (13.4%), Senegal and Tunisia (12.9%). Most of the continent clusters below 10%. Meanwhile, the United Arab Emirates hit 64% adoption, Singapore 60.9%, and even mid-tier digital economies like Norway surpassed 45%.
The adoption gap at a glance
End-2025 AI adoption. African observations highlighted.
Source: Microsoft AI Diffusion Index, end-2025
This is not merely a productivity story. The concentration of AI capability in the Global North has structural consequences.
Economic power
AI giants control frontier models, compute, and data infrastructure. Wealth from adoption concentrates where that stack already exists.
Brain drain
Researchers and engineers migrate toward centres with investment and opportunity, leaving developing regions perpetually behind.
Governance by others
Standards, procurement rules, and “responsible AI” are written where AI already exists. African governments adapt frameworks designed elsewhere.
Development slippage
Healthcare, agriculture, education, and governance are where AI could matter most under resource constraints. Without access, human-development gaps widen.
The Equiano shift: from safety to diffusion
The Equiano Institute was founded on a specific premise: responsible AI is not only about preventing bad futures. It is about ensuring that the good futures are distributed.
Over the past year, the global conversation has sharpened around this distinction. The African Union adopted its Continental Artificial Intelligence Strategy in July 2024 — the first comprehensive AI framework at continental level. More recently, the BRICS nations released a formal statement on global AI governance (summer 2025) that centred the concerns of low- and middle-income countries: access to data, environmental sustainability, decent work, and technological sovereignty.
These are no longer marginal voices. They are articulating that the problem is not only how AI is built — it is who gets to build it, and who benefits.
What this newsletter will track
State of AI Diffusion will report on six threads each quarter:
Access & Infrastructure
Data centres, GPUs, bandwidth, compute growth, and who is investing.
Adoption & Deployment
Where AI is actually used — healthcare, agriculture, finance — and what fails silently.
Local Capacity
Researchers, engineers, universities, and the shape of brain drain.
Governance Readiness
Auditing frameworks, impact assessments, procurement — homegrown or imported.
Economic Distribution
Whether AI creates wealth in the region or extracts it — and on what terms.
Innovation & Localisation
Locally developed models and regionally specific solutions versus imported defaults.
We will weave together data, policy analysis, case studies, and uncomfortable truths. Some of what we report will be optimistic — Africa’s AI talent is extraordinary, and there are pockets of remarkable innovation. Some will be sobering — infrastructure gaps are not closing fast enough, and open-weight models can mask deeper dependencies.
The stakes
AI diffusion is not a technical problem with a technical solution. It is political.
If the next decade sees AI’s benefits concentrated among the wealthy nations and corporations that controlled the infrastructure — while costs (labour displacement, environmental damage, systemic bias) are distributed to everyone — we will have missed a civilisational moment.
Africa is not a market to be tapped. It is a region of over 1.4 billion people, 54 countries with distinct economies and governance models, extraordinary talent, and immediate needs where AI could drive transformative change. Healthcare systems stretched thin could be augmented. Agricultural productivity could jump decades. Education could be personalised at scale.
But only if we ask the right questions first: Can Africa access this? Will it own it? Who decides?
What’s coming
In this inaugural issue, we are publishing three pieces:
- Introducing State of AI Diffusion (this post) — context and framing
- Global AI Diffusion: What the Data Shows — adoption gaps, infrastructure bottlenecks, and the widening North–South divide
- AI Diffusion in Low- and Middle-Income Countries — LMIC-specific challenges, agriculture and healthcare case studies, and an honest read on governance readiness
We will report quarterly. Each issue will include a data-driven report (with methods transparent), a deep dive on a sector or region, and analysis of policy developments that matter for the continent.
This newsletter sits at the intersection of AI capability and development. It is for policymakers charting adoption paths; researchers studying governance outside the Global North; investors looking for where innovation is actually happening; and practitioners deploying AI in constrained settings and learning what works.
The prior question is the point. Everything else follows from who gets to answer it.