Artificial intelligence is fundamentally disrupting the traditional economic development pathways that have guided emerging markets for the past 50 years, threatening to permanently widen the wealth gap between advanced and developing economies.
Analysis from Data Darbar reveals that AI is simultaneously undermining both major escape routes from poverty that nations have historically relied upon: low-cost manufacturing and offshore services.
For half a century, development economics operated under the convergence thesis. The theory that poorer countries would eventually catch up with richer ones by following proven pathways.
This theory emerged from the post-war, post-colonial era and suggested that newly independent nations could leverage the “advantage of backwardness” for rapid growth rates.
The US share of global GDP declined from 40% in 1960 to 26%, seemingly supporting the convergence narrative. However, almost the entire share ceded by Washington (13.8 percentage points) was captured by Beijing (12.5 percentage points), creating a bipolar economic order rather than the egalitarian global distribution globalists envisioned.
Historically, labor arbitrage coupled with weak currency enabled economies, particularly in Southeast Asia, to industrialize and build low-cost manufacturing bases, creating jobs at scale and providing forex buffers. This strategy powered the rise of countries like South Korea and China.
However, the model faces mounting pressures. According to the World Bank, 108 countries remain stuck in the middle-income trap, including economies that industrialized successfully as their cost advantage eroded faster than productivity advanced, with only 34 having crossed to “high-income” status since 1990.
Supply chain disruptions post-COVID revealed the risks of offshoring, while the broader pullback from globalization and tariff wars have pushed developed markets to consider reshoring. Most critically, as advancements in robotics materialize, the economics of labor arbitrage will change.
GenAI has upended the entire model as top AI models including Claude and GPT variants comfortably surpass what average workers produce, adjusted for price and delivery timelines.
Since March 2024, when Claude Opus 3 was released, India’s big 5 IT firms, i.e., TCS, Infosys, Tech Mahindra, HCL Tech, and Wipro, have seen a median stock decline of 18.6%, with noticeable declines in revenue and profit growth rates.
The conventional counterargument, that workers can simply upskill to use AI, provides little comfort. The question ultimately reduces to augmentation versus substitution: what share of jobs can gain productivity boosts from AI versus being replaced by it.
The escape route from both manufacturing and services challenges runs through the same door: frontier technology, but building that requires access to capital and compute at a scale increasingly concentrated in a handful of geographies. Financial returns from these investments would inevitably flow to advanced economies through capital income channels, widening the gap further.
The convergence thesis rests on two assumptions: that technology would diffuse outward over time, and that latecomers could always find something cheap to sell, but AI appears to be breaking both.
While some argue this represents just another tech cycle that will eventually correct, the immediate danger lies in market confidence itself. As long as enough people building and investing at the frontier of tech believe in AI’s potential, capital and talent will concentrate and the risk of divergence only amplifies, potentially reshaping global economic inequality for decades to come.