Glonce

AI is hyper-scaling the global digital divide, IEEE Spectrum essay argues

policyThe Glonce editorsIEEE Spectrum🔥67Glonce rating67Impact68Novelty52Relevance87Surprise48Players68Credibility82
AI is hyper-scaling the global digital divide, IEEE Spectrum essay argues

An opinion piece in IEEE Spectrum by Danica Radovanović, a digital-inclusion researcher and Senior Researcher at the Leibniz Institute for the Social Sciences, argues that AI is not closing the digital divide but hyper-scaling it, concentrating compute, skills and governance power in a small set of countries and firms. Citing Stanford's 2026 AI Index, the piece states the United States alone hosts more than 5,000 data centers, over ten times as many as any other single country, and, per World Bank data, accounted for roughly 87 percent of global exports of cloud computing and data-storage services in 2023. Because most AI workloads now run on cloud platforms rather than local infrastructure, the author argues this compute concentration becomes a concentration of dependency for most countries. On skills, the essay cites OECD figures showing only about 40 percent of adults across member countries have more than basic digital problem-solving skills, and that AI-related training itself is stratified by education: 36 percent of respondents with tertiary education reported AI-related training in the previous year, versus 18 percent of those with upper-secondary education. The piece then turns to two country cases. In South Africa, the Department of Communications and Digital Technologies released a draft national AI policy in April 2026 proposing new oversight institutions, then withdrew it days later after a journalist found at least six academic citations that did not exist, apparently AI-generated hallucinations; the minister called it "an unacceptable lapse." The author notes South Africa has since constituted a new AI panel with a chance to use the country's leverage. Indonesia is presented as a contrasting, more constrained case of public-sector agency: its National Research and Innovation Agency (BRIN) has built practical tools for underserved communities rather than frontier models, including an app that uses satellite data and machine learning to help artisanal fishermen locate fish, multilingual models trained on Indonesian and local languages such as Javanese and Sundanese, and government chatbots; in August 2025 Indonesia's Ministry of Communication and Digital Affairs released a national AI road map targeting the training of 100,000 AI-skilled workers annually. The essay also notes that in 2024 African ministers adopted a Continental AI Strategy and African Digital Compact, and that the April 2025 Global AI Summit on Africa in Kigali explored regional coordination, local-language models and open-source ecosystems as ways to reduce dependence on externally developed AI systems. The author's conclusion is not that AI development or cloud concentration is inherently bad, but that the more useful question is who gets to shape AI systems and priorities versus who can only adapt to systems built elsewhere.

Key facts

  • The US hosts over 5,000 data centers, more than 10 times any other single country, and captured about 87 percent of global cloud and data-storage export revenue in 2023, per Stanford's 2026 AI Index and World Bank data cited in the piece.
  • Only about 40 percent of adults across OECD countries have more than basic digital problem-solving skills, and AI-related training participation is split by education level: 36 percent among the tertiary-educated versus 18 percent among those with upper-secondary schooling.
  • South Africa withdrew its draft national AI policy days after its April 2026 release when a journalist found at least six fabricated academic citations, apparently AI-generated hallucinations; the minister called it "an unacceptable lapse."
  • Indonesia's BRIN has built AI tools for underserved communities, including a satellite-and-machine-learning app to help artisanal fishermen locate fish and multilingual models in Indonesian, Javanese and Sundanese, alongside an August 2025 national road map targeting 100,000 AI-skilled workers trained annually.
  • In 2024 African ministers adopted a Continental AI Strategy and African Digital Compact, and the April 2025 Global AI Summit on Africa in Kigali discussed regional coordination and local-language models as a hedge against dependence on externally built AI systems.

Why it matters

The essay's core claim is that AI is not a neutral technology landing on a level playing field: it is layering on top of decades-old gaps in connectivity, cloud infrastructure, skills and institutional capacity, and in the author's view amplifying rather than narrowing them. Because a handful of countries and firms control the compute that powers globally deployed AI systems, most other countries end up as consumers of imported models and standards rather than participants in deciding what those systems are built to do.

Who it affects

The piece frames this as a divide between a small group of technologically advanced states and firms that set the agenda, and everyone else, with South Africa and Indonesia offered as case studies of countries trying to participate without competing directly in the frontier-model race led by the US and China. It also touches on citizens on the losing side of the skills divide, who the author says are more likely to experience AI as an opaque system acting on them, such as the Dutch childcare-benefits algorithm scandal and Amazon's discontinued AI recruiting tool, both cited as prior examples rather than new reporting.

How to use it

The piece is aimed at engineers and policymakers rather than at consumers or investors; it argues for building AI infrastructure and governance processes that broaden participation from underrepresented languages, institutions and users rather than only adapting AI systems designed elsewhere. It points to Indonesia's road map (100,000 AI-skilled workers trained annually) and regional bodies like the African Union's Continental AI Strategy as concrete, if partial, examples of the kind of local capacity-building the author is calling for.

How solid is it

This is a signed opinion essay in IEEE Spectrum, not a reported news story, and its central claims rest on named sources: Stanford's 2026 AI Index, World Bank export data, and OECD survey data, plus two specific, checkable country episodes (South Africa's withdrawn draft policy, Indonesia's BRIN tools and road map). The author has a substantive background in digital-inclusion policy work, including with the UN's International Telecommunication Union, which supports the essay's on-the-ground framing, though the argument itself is her interpretation of the trend rather than new original research.

Risks and caveats

As an opinion piece, the essay's framing and emphasis are the author's own; the cited statistics describe structural gaps in compute, cloud exports and skills training but do not by themselves prove the essay's causal claim that AI is actively worsening those gaps rather than simply inheriting them. The South Africa case is a caution about a different risk entirely: a government AI policy document was withdrawn because it contained fabricated, AI-generated citations, a reminder that hallucinated content can reach official policy drafts if unchecked.

“an unacceptable lapse”

— South Africa's minister of communications and digital technologies, on the withdrawn draft AI policy