The Infrastructure of Intelligence Is the New Development Divide

For a generation, development policy has focused on closing the digital divide. That work remains unfinished. Roughly 700 million people still lack access to electricity, and more than 2.5 billion people remain offline. For them, the priorities are immediate: reliable power, basic connectivity, and affordable access to the digital world. That agenda is not going away. It remains foundational.

But even as we work to close that gap, a more consequential divide is emerging—one that will shape not just access to information, but access to economic opportunity, governance capacity, and national agency. The next development divide will not be defined only by who is connected. It will be defined by who has access to the infrastructure of intelligence—and on what terms.

The Cloud Was Never Weightless

Artificial intelligence is often described as something that lives in the cloud—intangible, weightless, available from anywhere. That framing obscures more than it reveals.

What we call “the cloud” is a physical system: semiconductor fabrication plants operating at atomic precision, hyperscale data centers consuming massive electricity, fiber networks carrying immense flows of data, and supply chains stretching across continents. AI does not float above the physical world. It is built within it.

And like all infrastructure, it concentrates—where capital is available, where energy is reliable, and where political and regulatory conditions are stable. That concentration is beginning to determine who participates in the AI economy—and who participates on someone else’s terms.

From Connectivity to Compute

In the last decade, the defining question was whether a country was connected. Today, the more important question is whether it has access to compute. Advanced AI systems require industrial-scale resources: specialized semiconductors, reliable electricity, high-capacity networks, and billions of dollars of investment. Many countries—despite strong human capital and growing digital sectors—lack one or more of these foundations. The result is a new form of inequality. Not digital exclusion, but computational dependence.

Two Paths to Participation

Yet dependence is not inevitable. Countries today face real choices about how they access and integrate into the infrastructure of intelligence.

One path is an open, interoperable model. In this approach, countries connect to global AI ecosystems in ways that preserve flexibility. Data and workloads can move across providers. Standards are shared. Governments can shape how systems operate within their legal frameworks. Dedicated or sovereign cloud environments can exist, but they remain connected to a broader, competitive ecosystem.

This model does not provide full technological sovereignty. But it preserves optionality—the ability to switch, adapt, and evolve. Over time, it tends to empower both citizens and institutions, supporting innovation, competition, and resilience.

The alternative is a more closed, vertically integrated model. Infrastructure, software, and services are bundled together—often delivered by a single provider or a tightly aligned ecosystem. These systems can be deployed quickly and at lower upfront cost. But they can also be difficult to unwind.

Over time, this can lead to vendor lock-in, where switching becomes costly or impractical. Control over data, standards, and system evolution may shift away from domestic institutions. In such systems, the same infrastructure that enables intelligence can also enable control. Not only over markets and data flows—but, in some cases, over the information environments and behavior of citizens.

Financing Access, Not Ownership

These choices are deeply connected to how countries finance access to artificial intelligence—not just how they build infrastructure. In practice, only a small number of countries will develop full AI stacks. The capital requirements, energy demands, and technical complexity place that option out of reach for most. Instead, most countries will access AI through segmented cloud models—drawing on infrastructure operated by global providers, sometimes located within their borders, sometimes regionally, and sometimes entirely offshore.

The real question is not whether they can build AI infrastructure. It is whether they can finance access on terms that preserve flexibility, governance, and long-term agency. This is where traditional development finance begins to strain.

Development banks were built to fund physical assets—power plants, ports, telecommunications networks. AI access is different. It is partly physical, but also contractual and institutional. It depends on long-term service relationships, data governance frameworks, and partnerships with private technology providers. That makes it harder to finance—and harder to evaluate.

The risks are different as well. A country may finance infrastructure tied to a single provider and find itself locked into a system that is difficult to exit. The most important questions—who controls data, who sets standards, who governs system behavior—are not easily captured in traditional lending models.

At the same time, the simplest projects to finance are often the most vertically integrated ones. They offer clear counterparties, turnkey solutions, and rapid deployment. More open, interoperable systems—those that preserve choice and adaptability—require more complex structuring, stronger institutions, and closer coordination between public and private actors. If development finance does not adapt, capital may flow toward the easier model. Not necessarily the one that best serves long-term national interests.

The Question of Agency

Artificial intelligence is becoming embedded across entire economies—healthcare, agriculture, finance, education, logistics. Countries that can deploy these capabilities effectively will see gains in productivity and resilience.

But access alone is not enough. The emerging divide is not simply about who uses AI. It is about who shapes it. Who sets the rules. Who captures the value. Who retains the ability to adapt as systems evolve.

In short, it is about agency within interconnected systems.

Rethinking Development Strategy

Development strategy must evolve to reflect this reality. Expanding electricity access and internet connectivity remains essential. But it must be paired with a new focus on how countries secure access to compute, how they finance infrastructure aligned with their interests, and how they integrate into global systems without surrendering flexibility.

The choice is not between isolation and integration. It is between forms of integration. One that empowers. One that constrains.

A Divide Redefined

The digital divide is not disappearing. It is being redefined. From access to the internet
to access to intelligence— and the terms under which that intelligence is delivered.

In line with my belief that responsibly embracing AI is essential to both personal and national success, this piece was developed with the support of AI tools, though all arguments and conclusions are my own.