Artificial intelligence has been a technology story for decades. That changed in 2025. When the United States placed China is top AI labs on the Entity List, when the European Union passed the AI Liability Directive, and when China deployed its first fully autonomous AI decision-making system in a military command structure, AI stopped being a software sector and became a domain of state power. The race for AI sovereignty has begun, and its outcome will determine which nations control the commanding heights of the global economy for the next fifty years.
The numbers are stark. Global AI infrastructure investment reached 780 billion dollars in 2025. Nvidia reported 115 billion dollars in revenue, driven almost entirely by H100 and B200 GPU demand. OpenAI is valued at 300 billion dollars. Anthropic at 75 billion. Mistral at 12 billion. The combined market capitalization of AI companies now exceeds the GDP of most nations. And the competition for the compute infrastructure that underpins all of it has become the defining geopolitical contest of our era.
The Compute Divide
The foundation of AI capability is compute. The most powerful AI models require thousands of specialized chips working in parallel, and the supply of those chips is tightly controlled by a small number of companies in a small number of countries. Nvidia controls 80 percent of the AI chip market. TSMC manufactures Nvidia is chips. ASML in the Netherlands makes the extreme ultraviolet lithography machines required to produce them. This supply chain runs through four countries and two companies, and every major power is now trying to break into it or control it.
The United States has responded with the most sweeping semiconductor export controls in history. The October 2022 chip rules, expanded in 2023 and 2024, restrict the export of advanced AI chips and the equipment to make them to China. Nvidia is H100 and later chips are effectively banned from the Chinese market. The goal is to prevent China from developing frontier AI capability that could be used for military applications. China is response has been to accelerate domestic chip development, pouring 150 billion dollars into semiconductor self-sufficiency programs through 2030.
The country that controls advanced AI chips controls the future of economic and military power. This is not about commercial competition. It is about which nations will have the capability to project power in the decades ahead.
— Former Google CEO Eric Schmidt, Congressional Testimony, February 2026
The Race for Model Superiority
Beyond hardware, the AI sovereignty contest is being fought on the model layer. The United States leads in foundation models, with OpenAI, Anthropic, Google DeepMind, and Meta all building systems of unprecedented capability. China has responded with its own frontier models, with companies like DeepSeek, Baidu, and Zhipu AI developing competitive systems. DeepSeek is R1 model, released in early 2025, matched GPT-4 performance at one-tenth the training cost, demonstrating that Chinese AI capability is not simply copying American breakthroughs but achieving independent innovation.
The European Union, meanwhile, finds itself in an uncomfortable position. It has strong AI research communities, world-class universities, and significant industrial applications, but no company at the frontier of foundation model development. Mistral, the leading French AI company, has a market valuation that is less than five percent of OpenAI is. The EU is response has been to focus on regulatory leadership, with the world is first comprehensive AI Act now in force, setting standards for AI governance that other jurisdictions are being forced to follow.
We are in a technology race, and the stakes are as high as they were in the nuclear age. Whoever leads in AI will have a decisive advantage in every domain of national power. This is not a race we can afford to lose.
— UK Prime Minister Keir Starmer, National AI Strategy Address, January 2026
The Data Sovereignty Question
AI models are only as good as the data they are trained on, and data sovereignty is emerging as the third front in the AI sovereignty contest. The European Union has passed the AI Act and the Data Governance Act, restricting the transfer of European citizen data to third countries for AI training purposes. China has its own data localization laws. The United States has no comprehensive federal data privacy law, which creates both a competitive advantage in terms of data availability and a diplomatic vulnerability in negotiations with the EU.
The emerging consensus among international law scholars is that AI training data constitutes a form of national resource, and that nations have the right to regulate its collection, processing, and transfer. This is transforming how countries think about digital trade agreements, creating new fault lines that go beyond traditional tariff and market access concerns into the fundamental architecture of how AI systems are built.
The Bottom Line
The AI sovereignty race is not a technology competition. It is a structural contest for power. The nation that leads in AI will have advantages in military capability, economic productivity, scientific research, and governance capacity that will compound over time. Those advantages will be difficult to reverse once established. The United States currently leads, but its advantage rests on a supply chain that is vulnerable to export controls, on a private sector that is answerable to shareholders rather than national strategy, and on a regulatory environment that has not yet decided whether AI is a commercial product or a national security asset. China is investing at a scale that has no historical precedent, with direct state funding, state-directed talent accumulation, and state-coordinated data collection. The outcome of this race will define the balance of power for the rest of this century. And the world has only just begun to understand what is at stake.
David Foster is a Senior Analyst for Media Hook, specializing in geopolitical analysis, economic trends, and the forces reshaping the global order.
