Artificial intelligence is often described as a technology that will spread rapidly because software can be distributed almost instantly. That view overlooks the physical and economic foundations required to use it. Before people can benefit from AI-enabled education, health care, financial services or government systems, they need dependable connectivity, an affordable device, sufficient data and the skills to navigate digital services safely.
The World Bank’s July 2026 review of the global internet gap shows how incomplete that foundation remains. Internet use has risen from 16% of the world’s population to more than 70% over two decades, but 2.2 billion people are still offline. In low-income countries, fewer than one in four people use the internet, compared with near-universal access in many wealthier economies.
The International Telecommunication Union’s ICT Development Index 2026 records continued progress across 159 economies, including strong improvements in low- and lower-middle-income countries. Yet it also finds persistent disparities in affordability, usage and connection quality. The next digital divide is not simply between having a signal and having none. It is between nominal access and connectivity capable of supporting meaningful participation in a modern economy.
Coverage is only the first layer
Mobile broadband coverage now reaches most of humanity. ITU data show that 3G or better is available to about 96% of the global population, while 4G reaches roughly 92%. Those numbers are important achievements, but they can create a misleading sense that the access problem is almost solved.
A network may cover a village while remaining too expensive for many residents to use regularly. A household may share one device with limited storage and an intermittent power supply. Congestion may reduce speeds when people most need the service. Language, accessibility and digital skills can determine whether an available connection becomes useful.
These constraints matter more as services become richer and more computationally demanding. A simple messaging application can operate over a weak network. Video learning, remote diagnostics, cloud business tools and AI assistants require greater bandwidth, lower latency and reliable data allowances. When public services move online without accounting for these differences, digitisation can shift inconvenience and cost onto the people least able to absorb them.
AI can widen an existing divide
AI systems are being integrated into workplaces, schools, customer service, agriculture and public administration. Organisations with strong infrastructure can combine these tools with cloud computing, proprietary data and skilled teams. Communities with weak connectivity may remain consumers of limited digital services—or be excluded entirely.
This produces a compounding effect. Better-connected regions gain access to new tools, generate more data, attract more investment and develop stronger digital capabilities. Poorly connected regions struggle to create the demand signals that justify additional infrastructure. The divide becomes an economic feedback loop rather than a temporary technology gap.
There is also a representation issue. Populations that participate less online create fewer digital records in local languages and contexts. AI systems built from uneven data may perform less effectively for those communities, reinforcing the same disparity at the application layer.
Meaningful connectivity is an operating system
The policy response must extend beyond tower counts. Affordable spectrum, competitive markets, shared infrastructure, reliable electricity, fibre backbones and local data centres all influence the quality and price of service. Satellite and direct-to-device networks can reach difficult terrain, but they need thoughtful regulation and business models that work for low-income users.
Demand-side measures are equally important. Device financing, targeted connectivity support, public access points and practical digital-skills programmes can convert coverage into participation. Online services should be designed for low bandwidth, inexpensive devices, local languages and people with disabilities. Offline and assisted channels remain necessary while access is uneven.
For businesses entering emerging digital markets, the lesson is similar. National connectivity averages are not enough. Product teams need to understand the actual devices, data costs, payment systems and network conditions experienced by customers. Lightweight applications and resilient service design can be competitive advantages, not merely concessions to limited infrastructure.
The last billions require a different model
The early expansion of the internet connected dense urban areas and customers who could afford service. Reaching the remaining population is structurally harder. Communities may be remote, incomes lower and infrastructure costs higher. Progress will depend on combinations of public finance, private investment, community networks and regulation that recognises connectivity as enabling infrastructure.
The AI era does not make this work less relevant. It raises the stakes. Intelligence delivered through the cloud cannot broaden opportunity when the route to the cloud remains unreliable or unaffordable. The most consequential digital strategy may therefore begin with an unglamorous question: who can connect consistently enough to participate?
Featured photograph: Sergej on Unsplash, used under the Unsplash License.
