Africa in the Age of Agentic AI: Competing Without the Bottom of the Stack

Over the past six months, I have been trying to make sense of what artificial intelligence really means for financial systems, particularly in Sub-Saharan Africa. The more I looked at it, the clearer one pattern became.

One conclusion has become clear: artificial intelligence (AI) is no longer a single breakthrough. It is a system that is layered across compute infrastructure, foundation models, data, applications, and now, increasingly, autonomous agents. And each layer behaves very differently depending on where you sit in the world.

This layered view of AI has been echoed in global discussions, including at the 2025 World Economic Forum, where NVIDIA CEO Jensen Huang described AI as spanning infrastructure, models, and applications at unprecedented scale.

Stanford University’s Human-Centred Artificial Intelligence (HAI) released the 2025 AI Index Report that further reinforces this reality. Leadership in AI is no longer defined by a single innovation, but by how many layers of the stack a country or company can shape and control (Stanford HAI, 2025; World Economic Forum, 2026).

This realisation led me to explore a deeper question: where does Africa fit into this emerging AI stack?

The Bottom of the Stack: Compute

The eighth edition, 2025 AI Index Report, is the most comprehensive to date and was published at an important moment, as AI’s influence changes across society, the economy, and global governance. The 2025 Stanford AI Index confirms a clear global trend: AI compute capacity and investment are increasingly concentrated in a small number of countries (Stanford HAI, 2025).

The United States leads in frontier model training and private AI investment, while China remains the second-largest AI power, investing heavily in semiconductor capability and model development. The European Union, the United Arab Emirates, and India are also advancing sovereign AI strategies, including large-scale investments in compute infrastructure and national AI capacity.

Global investment data shows that the majority of private AI funding remains concentrated in the United States and China, accounting for well over two-thirds of total investment (Stanford HAI, 2025; OECD, 2024).

Africa does not yet compete at this layer at scale, and realistically, it will not in the near term.

Sub-Saharan Africa accounts for a tiny share of global hyperscale data centre capacity, with South Africa leading regionally, followed by Nigeria and Kenya. However, GPU-intensive training infrastructure remains limited relative to global leaders (World Bank, 2024; ITU, 2023).

If AI leadership were determined only by compute, the outcome would already be decided, and Africa would already be out of the race.

Foundation Models: The Model Builders

The 2025 Stanford AI Index shows that the vast majority of frontier large language models are developed in the United States and China, with additional contributions from countries such as the United Kingdom, France, Israel, and increasingly the United Arab Emirates and India (Stanford HAI, 2025).

Africa has almost no meaningful representation at this layer today. Most advanced models used across the continent are accessed through global AI companies and cloud providers. In contrast, domestic frontier model development remains nascent, and talent pipelines are still evolving relative to global leaders.

The International Monetary Fund (IMF) has cautioned that artificial intelligence could widen global inequality if emerging markets remain primarily consumers of AI technologies rather than contributors to their development (IMF, 2024). That warning is not theoretical. It is already visible in how AI is being consumed versus built.

The strategic question is not whether Africa builds the largest model, but whether it builds intelligence on top of what already exists.

Data: Africa’s Structural Differentiator

This is where the story starts to shift. Africa is not compute-dense.
It is behaviour-dense, and that distinction matters more than it appears at first.

According to GSMA, Sub-Saharan Africa accounts for the majority of global mobile money accounts, with several markets processing billions of digital financial transactions annually (2024). Platforms such as M-Pesa, operating across multiple African countries, illustrate the scale and consistency of mobile-first financial behaviour across the continent. In countries such as Kenya, Ghana, Tanzania, and increasingly Nigeria and Ethiopia, mobile wallets often outnumber traditional bank accounts.

In this context, data is not extracted from legacy systems. It is generated through everyday financial behaviour. Africa’s mobile-first financial systems produce structured, high-frequency transaction data tied to P2P money transfer, airtime usage, microcredit, insurance premiums, remittances and merchant settlements.

This creates a fundamentally different data environment: financial data that is born digital rather than retrofitted from legacy infrastructure. In that sense, Africa’s digital finance ecosystem has already built a critical layer of the AI stack.

Applications: Where Value Emerges

According to the McKinsey Global Institute, generative AI could add between $2.6 trillion and $4.4 trillion annually to the global economy, with value realised primarily through real-world applications rather than models alone (2023).

In Africa, AI applications are emerging most visibly in sectors such as fintech, agriculture, health diagnostics, and telecommunications. Digital lenders already use behavioural scoring models derived from transaction histories and alternative data. Insurtech platforms price micro-policies based on usage patterns, while agricultural platforms apply machine learning to forecast yields and weather risks.

These are not frontier research breakthroughs. They are applied intelligence built on mobile-native systems, solving real constraints in markets where access, not sophistication, has historically been the binding constraint.

In many cases, these applications address inclusion gaps that advanced economies solved decades earlier.

Agentic AI: The Next Layer

The evolution of artificial intelligence is moving beyond generative models toward systems capable of executing tasks autonomously across complex workflows, often referred to as agentic AI (Stanford HAI, 2025). Discussions at global forums, including the World Economic Forum, increasingly point to this shift toward AI systems that can interpret intent, reason through objectives, and execute multi-step actions with limited human intervention (World Economic Forum, 2026).

Financial institutions globally are beginning to pilot AI-driven systems capable of underwriting loans, monitoring fraud, reconciling accounts, and interacting with APIs in increasingly automated ways. Advances in large language models and orchestration frameworks have significantly reduced the complexity of building such systems, enabling workflows that previously required extensive engineering effort.

Agentic systems change the equation. They reduce the importance of local frontier model development while increasing the importance of structured domain data and system integration. In finance-heavy, mobile-first economies, this shift is particularly significant.

Africa’s payment infrastructure, increasingly interoperable and real-time, creates an environment where intelligent systems could dynamically adjust credit limits, detect liquidity stress, or automate claims processing.

The constraint is not a behavioural signal. It is capital, access to computing, and a coordinated strategy.

Benchmarking the Continent

According to the Stanford AI Index 2025, the United States captured the majority of global private AI investment, followed by China, with countries such as Israel and the United Kingdom leading in AI startups per capita. India is expanding engineering capacity alongside public investment in AI infrastructure, while the United Arab Emirates is accelerating state-backed initiatives in compute and model development (Stanford HAI, 2025).

Africa’s AI investment volumes remain modest by comparison. Evidence from UNESCO and World Bank datasets points to lower AI talent density, research output, and R&D intensity across most Sub-Saharan African countries (World Bank, 2024).

But AI is a stack, not a single leaderboard. Africa may not control the foundational layers of compute and frontier models. Yet in mobile-native data systems and inclusion-driven applications, it has structural differentiation.

Countries such as South Africa, Nigeria, and Kenya are emerging as regional hubs for AI startups and innovation ecosystems (GSMA, 2024; World Bank, 2024). Ethiopia, while earlier in its AI journey, has established a national Artificial Intelligence Institute and is integrating digital systems into public service delivery. These are early signals, not yet dominance, but directionally significant.

The strategic question is whether the continent invests in the middle and upper layers, applications and agentic integration, before the global AI stack becomes structurally entrenched.

The Next Frontier

Global discussions on artificial intelligence increasingly point to one reality: the pace of model development is accelerating, and the window for strategic positioning is narrowing (Stanford HAI, 2025; World Economic Forum, 2026).

The rapid iteration cycles seen across leading AI labs make one conclusion unavoidable: waiting for sovereign compute parity is not a viable strategy for most emerging markets.

But this does not imply exclusion. It implies a different path.

Africa’s first digital revolution was connectivity. The second was mobile money. The third is now emerging, intelligent, autonomous financial systems built on top of these existing rails.

The defining advantage will not come from building the largest models, but from embedding intelligence into real economic systems, finance, agriculture, health and value chains, where data is generated continuously, and impact is immediate.

This is where the frontier shifts. From models to systems. From capability to application.

And this shift places a new premium not only on technology, but on execution capacity, the ability to design, deploy, and scale intelligent infrastructure in real markets.

Courses such as those offered by the Digital Frontiers Institute reflect this transition, equipping professionals with practical capabilities across digital money, mobile operations, remittances, agent networks, MSME finance, and agricultural digitisation, the very building blocks of Africa’s next financial systems.

The implication is clear: the next phase of AI leadership will not be defined by those who wait, but by those who build, fast, contextually, and at scale.

On that frontier, the race is not finished. It is just beginning.

By Yigermal Meshesha
General Manager (Intelligent Financial Services) at Kifiya Financial Technology PLC
Digital Frontiers Institute Alum and Community Member

 

Learn more about our Introduction to Digital Public Infrastructure (iDPI) course.

 

(Article also shared on Digital Frontiers on 11 May 2026)