The Geopolitics of AI

Geopolitics is the study of how geography and power interact — how the location of resources, populations, and infrastructure shapes the behavior of states. For most of history, the key geopolitical resources were land, oil, and access to trade routes. In the 21st century, a new resource has joined that list: artificial intelligence capability. Nations are competing for it, restricting access to it, and building alliances around it with the same intensity that earlier eras invested in controlling coal mines or controlling oil fields.

The Three Pillars of AI Power

To understand AI geopolitics, you must first understand what AI power is made of. It rests on three pillars: compute, data, and talent. Compute refers to the specialized hardware — primarily graphics processing units (GPUs) and AI accelerator chips — required to train large AI models. Training GPT-4 reportedly consumed tens of thousands of high-end GPUs running for months. A single cutting-edge AI chip, such as NVIDIA's H100, can cost $30,000. The semiconductor supply chain that produces these chips is extraordinarily concentrated: TSMC in Taiwan fabricates the most advanced chips; ASML in the Netherlands makes the extreme ultraviolet lithography machines needed to fabricate them; US companies design most of the chips and the underlying chip architectures. This concentration makes compute a major geopolitical chokepoint. Data refers to the large quantities of labeled and unlabeled information that train AI systems. Countries with large populations, sophisticated digital infrastructure, and a permissive (or state-controlled) approach to data collection have a structural data advantage. China's scale — 1.4 billion people generating data on tightly integrated digital platforms — gives it a significant data advantage in domains like facial recognition, payment behavior, and mobility patterns. Talent refers to the researchers, engineers, and practitioners who design, build, and deploy AI systems. AI talent is global in origin but concentrated in destination: the US attracts a disproportionate share of the world's top AI researchers, many of whom are themselves immigrants. In 2023, over half of US-based AI PhDs were born outside the US. Immigration policy is therefore AI policy.

The US-China AI Competition

The most consequential AI geopolitical rivalry is between the United States and China. Both governments have explicitly framed AI leadership as a national security imperative. The US has pursued a strategy of capability development combined with export controls. Since 2022, the US government has imposed increasingly strict restrictions on the export of advanced AI chips and chip-making equipment to China — specifically targeting Huawei and Chinese data centers. The logic: deny China the compute needed to close the gap in foundation model training, preserving a US lead in the most powerful AI systems. China's response has been an accelerated push for semiconductor self-sufficiency. The Chinese government has poured hundreds of billions of yuan into domestic chip development, seeking to reduce dependence on TSMC, ASML, and US chip designers like NVIDIA and Intel. Progress has been uneven — China's domestic chips remain several generations behind the frontier — but the long-term trajectory is toward greater independence. Beyond chips, the rivalry plays out in AI standards (both countries compete to set international AI standards through bodies like ISO and ITU), in AI-powered military systems (both are developing AI-enabled autonomous weapons, intelligence analysis, and logistics), and in AI diplomacy (both offer AI partnerships to developing countries, creating a competition for AI influence similar to Cold War-era competition for political allegiance).

It would be a mistake to frame this as a simple binary — 'US vs. China' — with all other countries as passive bystanders. The EU is a third major actor setting global regulatory norms. Japan, South Korea, and Taiwan are critical nodes in the semiconductor supply chain. India is courted by both blocs as a partner and is actively developing its own AI strategy. The United Kingdom, Canada, Singapore, Israel, and the UAE are significant AI hubs. The global AI order is multipolar, not bipolar, even if the US-China rivalry is the dominant axis.

Cooperation Alongside Competition

Competition is not the whole story. Parallel to the rivalry, there is significant international cooperation on AI — driven partly by the recognition that some AI risks are shared rather than zero-sum. The UK's AI Safety Summit at Bletchley Park in 2023 brought together representatives from 28 countries — including both the US and China — to discuss frontier AI risks. The resulting Bletchley Declaration acknowledged that 'there is potential for serious, even catastrophic, harm' from the most capable AI systems, and committed signatories to international cooperation on AI safety research. China's participation was notable: even fierce geopolitical rivals can recognize shared existential risks. Other cooperation forums include the Global Partnership on AI (GPAI), the OECD AI Policy Observatory, and bilateral research agreements between universities across competing countries. Scientific knowledge, unlike chips, travels more freely — though even research is increasingly subject to security scrutiny. The tension between competition and cooperation will define AI geopolitics for the coming decade. Nations want AI advantage but also recognize that a catastrophic AI failure anywhere — a grid-attacking cyberweapon, a pandemic-engineering AI — could harm everyone. This creates an unstable but real incentive for selective cooperation even between rivals.