- Mobile broadband now reaches 97% of the world, but only 59% actually use it, held back by five key barriers: access, affordability, knowledge and skills, safety and security, and relevance. Attention is shifting to AI before that gap is closed, and AI access depends entirely on closing it.
- Digital operators, companies built around connectivity that have grown into providers of payments, health and education services, have the reach and local relationships to help close it, but need policy support to do so.
- Concrete steps could help: spectrum policies that prioritise investment and reflect current market realities, smartphone financing, and a real role for operators in shaping national AI policy.
I remember when 4G still felt like a luxury in some of the markets we serve, something for capital cities, not villages. Mobile network operators fund 85 percent of all global investment in mobile internet infrastructure, $109 billion a year, and that money is why, per GSMA’s State of Mobile Internet Connectivity 2026 report, mobile broadband coverage now reaches 97 percent of the world’s population.
That investment was only ever half the job. The same research shows only 59 percent of people worldwide actually use mobile broadband, held back by five barriers: access, affordability, knowledge and skills, safety and security, and relevance. Their weight varies by market, but all five still matter.
I call what we’re building augmented intelligence, rather than artificial intelligence, because that’s a better description of what it actually does: tools that make an existing service, a doctor’s diagnosis, a teacher’s lesson, a shopkeeper’s credit decision, faster and better, without replacing the person doing it.
Five Familiar Barriers, Some Reappearing in Sharper Form
GSMA research identifies five barriers to mobile internet use: access, affordability, knowledge and skills, safety and security, and relevance. Look closely, and some of them, particularly affordability, skills and relevance, are reappearing around AI in a sharper form.
Affordability comes first. My colleague Vivek Badrinath, GSMA’s Director-General, said something worth repeating: AI is meaningless if people cannot get online, and the deeper risk is a divide between those who can afford to participate in the digital economy and those who cannot. Memory and chipset prices have nearly tripled over the past year, driven partly by global demand for AI infrastructure itself, pushing entry-level smartphones further out of reach. AI’s own infrastructure boom is pricing out the devices people would use to access it.
Digital skills come second, in a different form than before. Early mobile adoption asked little of a first-time user, dial, tap, scroll. Much of the world’s digital literacy effort over the past decade was built around exactly that simple model. AI asks something harder: knowing what to ask, and noticing when an answer is wrong. Most existing digital literacy programmes weren’t designed with that in mind.
Local-language relevance comes third, and may be the sharpest barrier of all. Most large AI models are trained overwhelmingly in English and a handful of major languages. A model that doesn’t work in Urdu, Bengali or Kazakh isn’t a lesser product for the hundreds of millions who use those languages daily; it offers them nothing at all. Basic internet access could survive, imperfectly, without local-language content. AI, built around conversation, cannot.
The five barriers are not new. What’s new is how AI can intensify existing challenges, including affordability, skills and relevance, at the exact moment the industry has a chance to close them instead.
Why This Is a Job for Digital Operators
We describe VEON as a services company built on top of telecom licenses held by our operating companies. Last year, for the first time, the number of customers using at least one of our digital services passed the number of customers on our telecom networks.
Scale matters here because of what it makes possible for one person. A missed loan isn’t an inconvenience to the shopkeeper who needed it, it can mean losing a season’s income. A delayed diagnosis, in a market where the nearest specialist is a day’s travel away, can mean catching an illness too late.
The companies building the largest AI models do essential work, pushing what these systems can do. But building a model and getting it to the people who need it most are different problems. Digital operators are closer to solving the second one: we have the network, local-language capability, and license obligations that make us accountable to governments. That position comes with responsibility: none of this works unless people trust how their data is used and know a human remains accountable throughout.
What Would Actually Help
Start with spectrum. Spectrum costs already represent a material share of operator revenues globally, around 7 percent, at a time when operators face pressure across the wider cost of doing business. Lower spectrum prices would leave operators with greater capacity to invest in their networks, expanding coverage, improving service quality and deploying new technologies. At the same time, rapid growth in mobile traffic means operators need more spectrum to keep pace with data demand. Spectrum should therefore be priced according to these market realities, rather than historical prices or outdated benchmarks.
Second, smartphone installment plans deserve wider backing. Given how sharply device prices are rising for reasons outside any customer’s control, this is no longer a nice-to-have.
Third, operators should have a seat in shaping national AI policy. We’re not simply funding the infrastructure, we see, in granular detail, where the gaps actually are: which households can’t afford a device, which languages a service fails to reach, which districts still have no signal at all. Policy written without that visibility risks solving the wrong problem well.
The Work That’s Still Open
Governments declared electricity and clean water universal necessities decades ago, and built policy to match: subsidies, obligations, universal-service rules. That declaration didn’t instantly deliver the outcome, hundreds of millions still lack both today, but it set the direction and the accountability that kept progress moving. Credit, diagnosis and education are each a foundation for a better life for all, the same way electricity and water are. AI is becoming the way many people will reach them for the first time, and they deserve the same kind of declared urgency, backed by policy, not just goodwill.
A decade ago, many people in the markets we serve had never used mobile internet on a device of their own. Today, far more do, because the industry worked to address the barriers holding back adoption, from access and affordability to knowledge and skills, safety and security, and relevance. Closing the AI divide doesn’t have to take as long, but only if operators, regulators and policymakers treat it with that same urgency, without losing sight of the connectivity work still unfinished beneath it.
The author is Group CEO of VEON, a Nasdaq-listed digital operator headquartered in Dubai, with operations in Pakistan, Ukraine, Kazakhstan, Bangladesh and Uzbekistan.