The next phase of the US-China confrontation will not be decided by tariffs. It will be decided by control. Officials from both countries have spent the past few days in New York discussing trade, investment, artificial intelligence and rare earths ahead of the Trump-Xi summit. The immediate objective is stability: extending the trade truce, implementing previous agreements and creating a framework for continued negotiations. A three-to-six-month extension remains possible. But the composition of the talks tells a more important story. The relationship is moving away from the old question of how much one country sells to the other. It is moving towards who controls the technologies, materials and infrastructure on which both economies depend. Artificial intelligence is the clearest example. Washington and Beijing have now agreed to continue a formal dialogue on AI risks, including the possibility of an incident-notification mechanism and emergency communications between the two sides. Another meeting is expected in Shenzhen within two months. That may look like cooperation. It is also evidence of how strategic AI has become.

The United States controls much of the most advanced semiconductor technology, software ecosystem and leading AI infrastructure. China controls something equally uncomfortable for Washington. It controls the physical supply chain underneath much of modern technology. Rare earths are the obvious pressure point. China accounted for about 94% of global production of rare-earth permanent magnets in 2024, according to the International Energy Agency figures cited in the material supplied. Companies such as JL Mag are moving beyond mining and refining into high-performance magnets, motor components and robotics, expanding the part of the value chain that would need to be replicated outside China. That changes the strategic equation. The West can build mines. It can subsidise refining. It can finance magnet factories. But China is not standing still while that capacity is being created. Its leading companies are scaling faster, moving into more sophisticated components and benefiting from industrial clusters developed over decades. The target is moving. JL Mag illustrates the problem. Its existing magnet capacity is already several times larger than that of emerging American competitors, and its planned expansion would widen that gap further. The company is also positioning itself in humanoid robotics, where demand for high-performance magnets could increase sharply. This is why rare earths have become leverage rather than merely commodities.

Beijing does not need to ban exports completely. Licences can be delayed. Volumes can be reduced. Specific categories can be restricted. Chinese exports of rare-earth magnets to the United States fell again in August ahead of the summit. The signal is enough. Washington understands that aggressive trade or technology restrictions can trigger retaliation in supply chains where American alternatives remain incomplete. Beijing understands that excessive pressure accelerates exactly the diversification it wants to prevent. That creates a strange form of mutual restraint. Neither side trusts the other. Neither side can fully disengage. The result is managed dependence. This also explains why the current negotiations extend beyond tariffs. Tariffs are visible. Supply-chain power is structural. A 20% tariff can be removed by executive decision. Replacing a rare-earth refining ecosystem, a semiconductor manufacturing chain or an AI infrastructure stack takes years.

Chinese companies are expanding production closer to Western customers. American companies are trying to rebuild strategic manufacturing at home. Governments are providing capital because markets alone may not accept the cost of duplicating existing Chinese capacity. Efficiency is no longer the only objective. Resilience has entered the equation. That makes the likely outcome of the Trump-Xi summit easier to understand. A truce is possible. Stability is possible. Selective agreements on AI safety, trade and critical minerals are possible. A return to the old relationship is much less likely. The United States and China are not negotiating their way back towards globalisation as it existed twenty years ago. They are negotiating the rules of a more controlled version of interdependence. The rivalry is therefore becoming more sophisticated. Tariffs were the first weapon. Export controls followed. Rare earths became leverage. AI is becoming strategic territory. The next phase will be about deciding which dependencies are tolerable, and which ones are too dangerous to keep. That is the real negotiation taking place in New York. Not how much America buys from China. Not how much China buys from America. But which country can afford to depend on the other for the technologies that will define the next decade.