The battle for global business and investment is entering a new arena. For decades, countries competed for multinationals with a familiar formula: lower taxes, skilled workers, predictable regulation, good infrastructure, and access to global markets.
Companies shifted talent and capital accordingly. The model helped Singapore become a regional base for multinationals, helped Switzerland attract global headquarters and high-value corporate activity, and helped turn Dubai into a magnet for international talent and regional business.
AI is adding a new layer to that contest. Access to advanced models is largely global, for now. A team in Singapore can often use the same model as a team in California. But deployment is local. A bank using artificial intelligence to make credit decisions needs clear audit trails, robust security, and certainty over where sensitive data resides. Increasingly, where AI can be deployed depends on four things: compute infrastructure, data sovereignty, AI governance, and continued access to frontier models.
That is starting to redraw the map of global business. Governments are competing to attract not only companies and talent, but the AI capabilities that could underpin future growth.
The geopolitical stakes are rising too. Last week, more than 20 countries backed a declaration calling for AI to remain under human control and proposed a global supervisory regime. At the UN General Assembly, President Donald Trump rejected the idea, arguing that the U.S. should keep building its lead over China. The debate reflects a broader shift: Countries increasingly see leadership in AI as central to national competitiveness.
And now, just as factories once clustered around rivers, ports, and coalfields, businesses are starting to bunch around AI. Clusters reinforce themselves. Capital attracts talent; talent brings capability; capability begets investment.
Countries are already competing across those dimensions, and in very different ways. Singapore is betting on trusted deployment. The city-state has spent years building AI governance frameworks, regulatory sandboxes, and AI Verify, a tool kit that lets companies test systems before rolling them out.
Bearing fruit
The strategy is bearing fruit: HSBC recently chose Singapore as the base for its Global AI Centre of Excellence, which will develop AI tools for use across the bank’s global business.
The Gulf has a different pitch to multinationals: Bring your AI workloads here, and we’ll provide the compute, energy, and data infrastructure to run them.
Saudi Arabia’s state-backed AI company Humain has unveiled data center projects and is targeting roughly 6 gigawatts of compute capacity over the coming decade or so. And the United Arab Emirates is building an AI computing cluster in Abu Dhabi rated at 1 GW, while its central bank works with the AI group Core42 on a “sovereign” cloud that keeps sensitive financial data inside the country.
China, meanwhile, is betting on diffusion, spreading the technology as widely as possible. By releasing increasingly capable open-weight models, Chinese labs such as Alibaba, DeepSeek, and Moonshot AI allow companies to download, adapt, and run models on infrastructure they control themselves. That shifts control. A closed-model provider can decide who gets access, but open weights, once downloaded, cannot be recalled.
Europe is going in yet another direction, betting that regulation can give it an edge. The AI Act, the EU’s comprehensive rulebook for the technology, is designed to make AI easier to trust. The test is whether it can do that without making AI harder to deploy. It reflects the broader tension at the heart of this week’s UN debate: how much oversight AI needs, and what that means for the pace and freedom of deployment. The EU is also investing in AI factories, and has a homegrown model developer in Mistral.
The U.S. starts from a position of strength. It already dominates much of the AI stack, from frontier models and cloud computing to advanced chips. Its strategy is to build on that lead by expanding data centers, energy and semiconductor capacity at home, while exporting American AI abroad.
But there is a catch. The same technologies the U.S. wants the world to adopt are increasingly being treated as strategic assets. In June, Washington briefly restricted foreign access to Anthropic’s Claude Fable 5 and Mythos 5 models.
Companies cannot ignore these shifts. They affect every major location decision. AI adds four variables to that decision—as noted: infrastructure, sovereignty, governance, frontier model access—and geopolitical risk can alter all of them
And the winner is?
Who will win? No country offers the ideal mix. So companies may need to spread their AI footprint across multiple locations. A bank might build AI capabilities in Singapore, keep European workloads inside the EU, run sensitive Gulf operations on sovereign infrastructure, and rely on U.S. frontier models where they make most sense. And turn to open-weight models where control over deployment and continued access matters most.
The questions become: Where should this capability live? Where should its data sit? Which models should it rely on? And how quickly can we move if the rules change? The difficulty for companies is that the answers keep changing, and the decisions are becoming harder to reverse. Geopolitics can shift overnight. Infrastructure takes years to build and decades to write off.
Build around a single model provider, cloud platform, or jurisdiction and you’re making a geopolitical bet, whether you realize it or not. Those bets are expensive to unwind, so today’s investment decisions could become tomorrow’s operating constraints. The hedge is optionality: the ability to switch models, including between closed and open ones, control of critical data, and having more than one infrastructure provider. That is the AI equivalent of supply-chain resilience.
None of that makes the old rules obsolete. Tax, talent, market access, labor costs, and capital still matter. Silicon Valley won’t lose its edge any time soon because another country makes AI easier to deploy. But the next set of economic hubs may emerge somewhere else.
The countries that make AI easiest to run will attract the investment, businesses, and talent that power the next phase of economic growth. The rest will watch that growth happen elsewhere.