Artificial Intelligence and the Acceleration of Global Economic Convergence: Whether inexpensive intelligence can compress capability gaps between advanced and emerging economies

Abstract

Artificial intelligence has undergone a historic price collapse—“the cost of achieving GPT‑3.5‑class performance fell from $20 per million tokens in 2022 to $0.07 by late 2024”—driven further by the diffusion of open‑weight models that make frontier‑adjacent capability accessible to anyone with basic connectivity. This creates a new form of digital capital whose marginal cost approaches zero, raising the possibility that emerging economies could leapfrog traditional development barriers.However, the report shows that falling software prices do not eliminate the structural constraints that shape technology diffusion. Compute capacity remains extraordinarily concentrated—“roughly 90 percent of the compute inside the world’s leading AI supercomputers sits in two countries”—and energy, investment, and data‑center infrastructure follow similar gradients. Meanwhile, adoption gaps mirror earlier digital divides: coverage is not the same as usage, and low-income countries face persistent barriers in devices, literacy, and institutional readiness.To explain these dynamics, the report extends the H Heuristics D‑Coefficient (absorptive capacity) and P‑Integral (cumulative burden) into a Compound Convergence Multiplier, showing that convergence occurs only when absorptive capacity rises alongside falling AI costs. The proposed LEAP Framework—Localize, Enable, Adopt, Propagate—identifies the four complementary pillars required for countries to convert inexpensive intelligence into real economic capability.Case studies from India, Kenya, Rwanda, Ethiopia, Uganda, and Nigeria demonstrate that meaningful gains are already possible even without frontier compute, particularly in agriculture, health diagnostics, education, and financial inclusion. Yet they also reveal how lagging pillars—especially compute and governance—limit the multiplier.The report concludes that inexpensive intelligence is a necessary but not sufficient condition for global convergence. The decisive variable is not the price of AI itself, but whether governments build the complementary infrastructure, skills, and governance needed to absorb it. As the report states: “Inexpensive intelligence removes one historic barrier… while leaving the older barriers of power, skills, and governance largely intact.”