by Zhao Xiaopeng

The first three industrial revolutions left much of the Global South behind. Justin Yifu Lin argues that the fourth need not - for one reason: this time, the frontier technology is within reach.

“If the Global South cannot capture this opportunity and use AI in production, the gap between the North and the South will widen further,” the Peking University economist and former World Bank chief economist told China Economic Net after the 18th BRICS summit in New Delhi.

AI access and co-operation were also on the BRICS agenda. At the summit, Chinese President Xi Jinping called on BRICS to become “pioneers of the times” in four areas — innovation-driven development, peace and stability, mutual learning among civilization, and reform of global governance. A day later, China followed with five initiatives on AI co-operation, among them a pledge to take the lead in establishing a BRICS AI open source community.

For Lin, wider access to AI matters because it could allow developing economies to apply advanced technology in industries suited to their own conditions — lowering one of the barriers that left them behind in previous industrial revolutions.

AI divide

The infrastructure underpinning AI remains deeply unequal. The infrastructure underpinning AI remains deeply unequal. Upper-middle- and high-income economies held 95 per cent of global co-location data-centre capacity in June 2025, while lower-middle- and low-income economies accounted for only about 5 per cent, according to the World Bank.

A similar divide is visible in adoption. On Microsoft data, generative-AI use in the second half of 2025 reached 24.7 per cent of the working-age population in the Global North, against 14.1 per cent in the South.

Yet the picture is more complicated than a simple North-South split. More than 40 per cent of ChatGPT’s global traffic in mid-2025 came from middle-income countries, led by Brazil, India, Indonesia and Vietnam.

Demand, in other words, is already substantial even where infrastructure and access lag behind.

That gap between demand and capacity is central to Lin’s argument. He contrasts two paths for frontier AI development.

The dominant U.S. commercial model, in his account, relies heavily on capital, advanced chips and computing power, which requires higher investment and returns. While China has pursued what he describes as a more software-intensive and talent-driven route, with open-weight releases lowering the cost for others to access and adapt advanced models.

“That means other countries can use AI technology to improve their productivity, and they don’t have to fall behind,” Lin said.

DeepSeek-R1 and Alibaba’s Qwen3 are examples: their weights can be downloaded, run and adapted, subject to their respective licences.

The result is an ecosystem rather than a series of one-off releases. Hugging Face hosts more than 151,000 models derived from Qwen, illustrating how open-weight models can be reused and adapted by other developers.

A developer in Lagos, São Paulo or Jakarta can download an advanced model and adapt it to local needs without negotiating a conventional technology-transfer agreement.

That portability is central to what Lin means by an open system — and why he thinks it can travel.

“The Chinese way is not only good for China,” he said. “It is good for the Global South, and in the end it may also benefit the Global North.”

A potential latecomer advantage

Affordable access is only the first half of the argument. The second is that latecomers face fewer legacy costs.

“When a developing country enters a new sector, the sector is totally new to it, so it can apply the latest version of technologies,” Lin said. “When they enter a new sector consistent with their comparative advantages, they can immediately apply AI.”

Early deployments offer a glimpse of how that could work, though still at small scale. In Uganda, Sunflower, an agricultural chatbot built by Kampala-based non-profit Sunbird AI on Qwen open-weight models, provides farming and food-security advice in dozens of Ugandan languages.

But access alone does not create a latecomer advantage. For Lin, turning open models into productivity depends on a more familiar development principle: comparative advantage..

In his New Structural Economics, economies develop by moving into industries consistent with their comparative advantages, which evolve as their endowments — labour, skills, capital and resources — change.

AI does not suspend the rule. But it can lower some of the barriers to technological upgrading.

That leaves a crucial gap between access to a model and its productive use. Bridging it, Lin argues, ultimately has to happen locally.

“When each country wants to use AI to improve productivity in a sector where it has a comparative advantage, it has to do that by itself,” he said. Scale, market access and production processes differ everywhere. Countries need “local talent to transform the technology into local applications,” Lin added.

Weight, and voice

For Lin, the question of who benefits from technological change is part of a broader imbalance between the Global North and South — not only in access to technology, but also in influence over the institutions and ideas that shape development.

AI is becoming an object of that contest. The BRICS AI initiatives are intended to give developing countries a greater role in shaping how the technology is used and governed.

WAICO, the World AI Co-operation Organization, is China’s bid to build a multilateral forum for AI governance; the BRICS application center and open source community give developing countries a stake.

“Currently, the distribution of voting power is very unfair — especially some countries that have veto powers, even though their weight in the economy is very small,” he said of the IMF and the World Bank.

The shift in economic weight is already substantial. The 11 BRICS economies accounted for about 41 per cent of global GDP in purchasing-power-parity terms in 2025, compared with 28 per cent for the G7, according to IMF data.

“We need to have a new global governance that reflects the successful experiences of the Global South — not the ideas prevailing in the Global North imposing an unequal system on the Global South,” he said.

Beyond technology

Yet for Lin, the deeper question goes beyond technology and institutions. Asked which of the four “pioneers” he would choose if he could pick only one, Lin chose mutual learning among civilizations — over artificial intelligence or any of the others.

“Fundamentally, ideas are the most important thing for guiding our efforts,” he said. “Practice can differ, but the basic principle should be the same: make yourself competitive.”

For much of the postwar period, development ideas flowed largely from developed to developing economies. Lin argues that learning should now run both ways.

“Now we need to learn from each other within the Global South,” he said. “And the North can also learn from the Global South.”

If that learning takes hold, the fourth industrial revolution could give the Global South something the previous three denied it: not just access to the technologies reshaping the world economy, but a greater hand in shaping the rules and ideas that govern it.