The first phase of the AI revolution belonged to frontier models. A handful of companies invested hundreds of billions of dollars in chips, data centres and increasingly powerful systems, and the global conversation followed: which lab has the strongest model, which country will lead.

While those questions matter, as AI matures, a different question is becoming just as important: which countries will translate this technology into real economic and social progress? The answer will not be decided in the labs building frontier models. It will be decided in frontier countries applying them.

History’s pattern here is stark. Economists Diego Comin and Bart Hobijn, who tracked technology diffusion across countries and centuries, found that steam-era technologies took over a hundred years to reach much of the world. Electricity took roughly half a century. The internet cut the lag to under a decade. AI has done something no previous technology managed: it collapsed the lag to nearly zero and then inverted the curve.

OpenAI’s own usage research found ChatGPT growing fastest in low- and middle-income countries, a reversal of every previous technology’s adoption pattern. For the first time, a general-purpose technology is arriving in the developing world at the same moment it arrives in Silicon Valley and not a generation late.

ChatGPT is growing fastest in low- and middle-income countries, a reversal of previous technology’s patterns

That simultaneity is Pakistan’s opening, but it will not stay open. A country already one generation behind technologically risks falling another generation behind if it manages this transition poorly; but if it manages it well, AI can accelerate progress across classrooms, clinics, farms, factories, and public institutions at a pace no previous wave of technology could match.

Frontier models may define the leading edge of capability; frontier countries can define the leading edge of application, and it is at the application layer that most of AI’s economic value will ultimately be created.

The sceptic’s objection is fair: a country managing energy constraints, fiscal pressure and the emigration of skilled talent is an unlikely candidate to lead anything in AI. But that is precisely why the adoption story matters more for Pakistan than the compute story. Leading at the application layer does not require winning the chip race. It requires connectivity, digital rails, relevant skills and the will to deploy, and on those fronts, the groundwork is further along than the sceptics assume.

Pakistan’s technology sector has built real momentum over the past decade. Systems Limited competes globally, reflecting the depth of local engineering talent. IT and IT-enabled services exports reached approximately $3.8 billion in the first ten months of FY26, while freelancer earnings are expected to approach $1bn.

The physical foundations are being laid too: the launch of Sky47, the country’s first AI factory, marks a first step towards sovereign AI compute, alongside continued investment in connectivity and cloud infrastructure.

Pakistan brings two structural strengths to this effort. The first is demographic: more than 60 per cent of its 250 million citizens are under thirty, and each year hundreds of thousands of graduates enter the workforce, a growing share of them trained in science, technology, engineering and computing.

This is one of the world’s largest populations positioned to adopt AI at scale and, as the ChatGPT data shows, young users in lower-income countries are exactly who is driving global AI adoption today. The second is the digital foundation built over the past decade. Connectivity, digital payments and digital identity are the rails on which AI will travel, and Pakistan has spent ten years laying them.

The third advantage is context. Across healthcare, education, finance, agriculture, commerce and public services, millions of daily digital interactions reflect how Pakistan actually learns, works, trades and communicates. As AI becomes embedded in everyday life, solutions grounded in these local realities will become a genuine source of competitive advantage.

Language makes the point concretely. Urdu, Punjabi, Sindhi, Pashto, Balochi and Saraiki collectively represent tens of millions of speakers, yet remain underrepresented in global AI development. And with millions of Pakistanis still on feature phones, broad access matters as much as model sophistication. The real measure of AI’s impact will not be benchmark scores; it will be whether a farmer, a nurse or a shopkeeper can use it in their own language, on the device they already own.

The policy environment is moving in the right direction. The National AI Policy and National AI Innovation Framework recognise AI as a competitiveness, governance and development priority, and their emphasis on talent, research and industry-academia collaboration reflects an important truth: long-term leadership will depend as much on people as on technology. What is needed now is execution with sustained investment in skills, reliable digital infrastructure, and a regulatory environment that lets builders build.

The AI era will be shaped both by the countries that build the most powerful models and by those that apply them most effectively. Every previous technological revolution reached countries like Pakistan decades late, and those decades compounded into the gaps we live with today. This one has arrived on time. Somewhere in the world, a frontier country will demonstrate what it looks like to turn that head start into opportunity, productivity and growth at scale. Pakistan has the people, the rails and the urgency to be that country — if it moves now.

The writer is the CEO & President of JazzWorld, as well as Chairman of Mobilink Bank & JazzCash.

Published in Dawn, The Business and Finance Weekly, August 17th, 2026