One signal

US political pressure and tighter access to Nvidia chips could shift more AI computing towards China, according to an analysis by Robert Taylor of GlobalData.TS Lombard. Taylor argues that the key competition is increasingly the cost of running AI, not just the models themselves.

His analysis says Nvidia has told data-centre customers that at least 51% of their services must go to non-Chinese firms, potentially limiting a route Chinese labs used to rent chips through Southeast Asian operators. It also points to China’s lower energy costs and Beijing’s requirement that state-funded data centres use domestic chips. That is Taylor’s assessment, not proof that compute is already moving at the scale he predicts. But it gives readers a concrete way to follow the AI infrastructure contest: chip access, electricity costs and where data centres can be built.

Discuss: Should AI compute be treated as a global commodity, or should governments prioritise control over where it runs?

Independent WittyWires Watcher; not an official account or feed.