US vs China vs Rest
The agent economy splits cleanly: the US owns the value-capture layers (models, cloud, chip design ~83-90%), China competes on price under a 7nm ceiling, and the deepest leverage sits upstream with allies: TSMC (Taiwan) and HBM (Korea). Estimates marked ~.
Who is ahead in AI, the US or China?
The US leads the layers where the money is made: frontier models, hyperscale cloud and chip design. China competes hard on efficiency and price, and its open-weight models are widely used, but it is limited by restricted access to the most advanced chips.
Why can't China buy the most advanced AI chips?
US export controls restrict the sale of top-end accelerators to China, and also the equipment needed to manufacture them. The response has been domestic chip programmes and unusually efficient model designs that get more out of less compute.
Why is Taiwan so important to AI?
Because TSMC, based there, fabricates almost all the leading-edge chips that AI runs on. The designs come from the US, but the manufacturing is concentrated on one island, which is why it is treated as a strategic risk rather than a commercial one.
What is ASML and why does it matter?
A Dutch company that builds EUV lithography machines, the only tools capable of printing the smallest chip features. It is the sole supplier in the world. No leading-edge fab anywhere can be built without its machines, which makes it the single narrowest point in the whole chain.
What is sovereign AI?
A country building its own models, chips or data centres so it is not dependent on foreign infrastructure. It costs more than renting capacity abroad. The point is control: over data, over continued access, and over who sets the rules.