How will AI reshape the global balance of power between the US, China, and the EU over the next decade
Who Gets to Run the World's Most Powerful Computers?
Based on analysis of a 119-node, 490-edge knowledge graph mapping the forces shaping AI geopolitics through 2035.
What This Is About
Imagine that over the next ten years, AI becomes as important to national power as electricity or nuclear weapons — not as a weapon itself, but as the engine behind economic growth, military capability, and global influence. The question then becomes: who controls the AI? Who makes the rules? And what happens to everyone else?
A detailed map of how different forces interact — countries, technologies, policies, companies, and feedback loops — was built and analyzed as a structured knowledge graph. What follows is what that map shows, explained plainly.
The Two Things That Matter Most
The map has 119 concepts and 490 connections between them. Two nodes have more connections than anything else, each linked to 42 other things at the highest importance level.
The first is who controls the physical hardware — the specialized chips and data centers that run AI. Think of this as controlling the factories that make the engines of the digital economy. Whoever can build or block access to the best chips holds enormous leverage.
The second is the fact that the US, China, and the EU cannot agree on any shared rules for AI. This isn’t a temporary disagreement — the map treats it as a structural outcome, meaning it’s not something that happens because of one bad negotiation but because the three blocs have genuinely incompatible goals and values around AI development.
These two things are treated as equally central in the map. That’s a claim worth pausing on: the political breakdown is not shown as a side effect of the hardware race, and the hardware race is not shown as a side effect of the politics. They are co-equal. The physical layer and the political layer are locked together.
The US Strategy That Keeps Getting Poked Full of Holes
The US has a policy of restricting what advanced chips and chip-making equipment can be sold to China. The idea is to slow China’s AI development by limiting access to the best hardware. On the map, this policy shows up as one of the most-connected nodes.
But it is also the single most-undermined concept in the entire graph. At least thirteen separate mechanisms are shown pushing against it or punching holes through it. Here is a simplified tour of a few:
The efficiency problem. When China couldn’t get the best chips, Chinese researchers built AI systems that work surprisingly well with cheaper, older chips — most visibly through the approach taken by the DeepSeek models. This created a paradox: the restriction designed to keep China behind instead created pressure to innovate around the restriction, which produced techniques that are now available to everyone, including through open-source releases.
The open-source problem. When you publish an AI model openly — meaning anyone can download and run it themselves — export controls become almost irrelevant. You can’t restrict a software download with a chip embargo. China has released capable open-weight models, and the map shows this directly undermining the control strategy.
The financial plumbing problem. US export controls partly work because the US can threaten to cut off access to the global financial system (which runs through US-controlled infrastructure). China has been developing alternative payment systems that don’t depend on that infrastructure. The map shows this eroding one of the enforcement levers for the chip restrictions.
The map does not say the export control strategy has failed. It records that it is under simultaneous pressure from many directions, which is a different claim.
The Vicious Cycles
Several concepts in the map form closed loops — situations where one thing causes another, which causes the first thing to get worse, which causes the second thing to get worse, and so on.
The race-governance spiral. The less international agreement there is on AI rules, the stronger the incentive for each country to race ahead and win before the rules catch up. The stronger the race incentive, the less likely any country is to slow down for governance discussions. This loop has no interrupting mechanism in the map. It just spins.
The Taiwan loop. Taiwan makes the most advanced chips in the world, and its continued independence is partly protected by how dependent everyone is on its chip factories. But as AI-driven military technology develops — particularly autonomous drones and AI-guided weapons — the possibility of a conflict over Taiwan becomes more discussable. If that conflict happens, global chip production collapses, which is itself a massive AI power event. The map shows these two ideas pushing each other in a reinforcing loop, both edges at the highest weight level.
The EU’s self-defeating regulation loop. The EU has tried to use its large market as leverage: set strict AI rules, and companies worldwide will follow them rather than lose access to European customers. This worked with data privacy (GDPR) and with product safety. The map shows the EU attempting the same move with AI.
But the map also encodes a problem: the EU’s strict rules push talented AI researchers and companies to move to the US or elsewhere, where there are fewer restrictions. This drains the EU of the AI capability it would need to actually compete — and an EU that imports all its AI rather than building its own may eventually lose the credibility to set rules that anyone has to follow. The very strategy meant to establish sovereignty keeps undermining the capability base that makes sovereignty meaningful.
The Surprising Connections
A few connections in the map are not obvious and worth highlighting.
EU restrictions on US AI may open doors for China. When EU regulations make US AI products harder to use in Europe, the market space doesn’t disappear — it becomes available for other providers. Chinese open-source AI models, which the EU might treat as less politically threatening than dominant US platforms, can fill that gap. The intended effect (reducing dependence on US tech) produces an unintended effect (facilitating Chinese AI penetration into European markets).
India’s middle-ground strategy depends on the thing it’s avoiding. India has been positioning itself as a third option — neither fully in the US camp nor the Chinese camp — for developing countries choosing which AI systems to adopt. But this positioning only works if US restrictions on China are tight enough to prevent China from simply dominating everywhere. If those restrictions erode (see the vicious cycle above), India’s space to maneuver shrinks. India’s independence is structurally dependent on US policy working.
The main US-led military alliance structurally excludes the country that matters most for the developing world. AUKUS — the security pact among the US, UK, and Australia — is shown as deliberately excluding India from its AI-and-defense elements. But India is also the node most connected to the developing world’s non-alignment posture. The US military architecture and the US soft-power architecture are working at cross-purposes in the map.
Gulf oil money could be a wild card for chip independence. The map shows Gulf sovereign wealth funds — government investment pools from countries like Saudi Arabia and the UAE — investing heavily in AI infrastructure. Crucially, this investment isn’t tied to either US or Chinese chip supply chains. If Gulf-funded AI clusters reach real scale using chips from Japan, South Korea, or other suppliers, they could represent a third pathway that the current US-China framing doesn’t account for.
The Developing World Is Not a Settled Question
A significant portion of the map concerns what is called the “Global South” — countries across Africa, Latin America, South and Southeast Asia, and elsewhere that are not themselves major AI producers but will be major users and, crucially, are being courted by both the US and China.
China has been exporting AI-powered surveillance systems, smart city infrastructure, and telecommunications equipment to many of these countries through programs connected to the Belt and Road Initiative. Countries that adopt this infrastructure become dependent on Chinese systems, which shapes their political alignment over time.
But the map does not show this as a resolved contest. The developing world is shown resisting full alignment with either side, and several mechanisms work against China’s influence in this space — including India’s alternative model of building open, public digital infrastructure rather than proprietary Chinese or US systems.
The map specifically encodes a time limit here: the window for developing countries to make these choices without being locked in is roughly 2027-2035. After that point, infrastructure decisions made in this period become very difficult to reverse.
The Thing Nobody Is Doing Anything About
The map contains a concept called the “AGI governance vacuum” — the absence of any agreed international framework for managing the development of potentially transformative AI. Multiple nodes in the map point toward this vacuum and amplify it. Zero nodes in the map show anything closing it or shrinking it.
This is a structural observation, not a prediction: within the map as constructed, the vacuum is shown only as a destination, never as a recoverable state.
Bottom Line
Five things stand out from the map’s structure:
1. Hardware and politics are locked together. Who controls the chips and who controls the rules are treated as equally foundational — neither causes the other, and neither can be solved independently.
2. The US export control strategy is under more simultaneous pressure than any other major node in the graph. The map does not conclude it has failed, but it records that it faces more undermining forces than any comparable mechanism — from efficiency innovation, from open-source releases, from alternative financial infrastructure, and from rare earth leverage.
3. The EU’s regulatory strategy contains its own failure condition. The Brussels Effect works when the EU is a market worth entering. If the EU’s competitiveness keeps declining — partly because its own regulations push talent away — the credibility of its regulatory leverage declines with it.
4. The developing world is the most genuinely unresolved contest in the graph. Unlike most other dynamics, which show clear directional pressure, the competition for Global South alignment is shown as actively contested, with no mechanism resolving it before the lock-in window closes.
5. The governance race-to-the-bottom loop has no brake. The feedback loop between weak international governance and strong first-mover incentives is shown as a closed, uninterrupted reinforcing cycle. The map offers no mechanism that slows or reverses it.