The easiest part of the AI boom to see is the interface. A question appears in a box. An answer comes back. A new model arrives and the internet debates whether it is smarter than the previous one.
The hardest part to see is the machinery underneath.
Every AI service relies on chips, data centres, power, cooling, networks and financing. As demand grows, the companies and countries able to secure those foundations gain an advantage that cannot be copied with a clever prompt.
That is why this week’s global technology conversation deserves to be read through an infrastructure lens. Web Summit is taking place in Lisbon, bringing founders, investors and technology companies into the same conversation. But the larger story stretches from US technology firms and Asian chip supply chains to European energy systems and the ambitions of emerging economies.
The AI race is no longer simply about who has the best model. It is also about who can afford to run one, who can build the physical capacity to support it and who sets the terms of access for everybody else.
This matters because the economics of AI are uneven. A small startup can experiment with an API. A company trying to serve millions of customers may face substantial ongoing compute costs. A research institution without reliable access to computing power can struggle to compete even when its ideas are excellent.
The imbalance is not unique to Africa, but its consequences can be especially sharp here.
Imagine an African health-tech company that wants to build a system capable of working in multiple local languages and low-connectivity environments. It needs data, technical expertise and regulatory trust. It may also need computing resources priced in foreign currency. Its challenge is not just to invent something useful. It must build under conditions set partly by industries and institutions elsewhere.
This is why discussions about semiconductor supply chains and cloud concentration are not distant geopolitical theatre. They can shape what local developers can afford to create.
Still, infrastructure should not become an excuse for fatalism. Africa does not need to manufacture every advanced chip to participate meaningfully in the AI economy.
There are opportunities in reliable power, connectivity, regional data centres, cybersecurity, data stewardship, model evaluation, language technology, sector-specific products and technical maintenance. None is glamorous in the way a viral chatbot launch is glamorous. Many may be more durable.
The uncomfortable question is whether we are preparing people for those opportunities.
Our technology conversations still lean heavily towards learning to use the newest tool. That is useful, but incomplete. We need people who understand systems engineering, energy, security, data governance and the economics of operating technology at scale.
We also need to be more demanding of announcements. A proposed data centre is not the same as a functioning one. A partnership is not the same as locally retained expertise. A national AI strategy is not proof that schools, researchers or small businesses can access the resources they need.
When the headlines from Lisbon fade, those distinctions will remain.
The world is investing in the physical foundations of AI. Africa’s question is not merely whether it can join the conversation.
It is which parts of that foundation we intend to build, own, maintain and improve ourselves.