KAIST blueprint urges ‘Bridge Power’ bloc for frontier AI

This report is one of the most concrete attempts yet to sketch an alternative to a US‑China duopoly over frontier AI. By foregrounding hard numbers on compute—75% of global capacity in the US, 15% in China—the authors force mid‑sized economies to confront a simple reality: going it alone on trillion‑parameter models and the data centers behind them is no longer realistic. A CERN‑style consortium lets them share capex, talent pipelines and data governance frameworks rather than each building a miniature, under‑scale OpenAI or DeepMind.

For the race to AGI, a "Bridge Power" bloc could accelerate progress by bringing more pluralistic research cultures and safety perspectives into the frontier club, while also diluting the dominance of a handful of US tech giants. If Korea, Europe and Singapore co‑fund large‑scale clusters and foundational models, they gain leverage over standards for evaluation, red‑teaming and compute governance. That, in turn, could make global coordination on AGI safety less like pleading with two superpowers and more like negotiating among a coalition.

The open question is execution: multilateral science projects are famously slow and bureaucratic. But even a partial success—shared sovereign AI infrastructure among a dozen democracies—would meaningfully reshape who gets to set the norms for AGI development.