By South China Morning Post | Created at 2026-06-11 18:15:52 | Updated at 2026-06-11 23:46:19
5 hours ago
NVIDIA’s top executive for Latin America denied on Wednesday that the region has served as a corridor for restricted chips into China, weeks after Anthropic, the American maker of the Claude AI models, alleged that Chinese labs had relied partly on smuggled processors to drive recent advances.
Speaking at Web Summit Rio at a moment of
intensifying rivalry between Washington and Beijing over artificial intelligence, Marcio Aguiar acknowledged that the pressure on export controls is real enough to reach his sales desk, even if the suspicious orders come from elsewhere.
“Companies appear, for example, in countries with which we have never had commercial relations, and they want to buy large quantities,” Aguiar said.
“So we ask, what are you going to buy this for? Where is the data centre? I need the documentation.” When the answers fall short, he said, the company walks away. “For several reasons, we do not sell.”
The accusations he was responding to came from a paper Anthropic released in mid-May, while US President Donald Trump was in Beijing meeting with Chinese leader Xi Jinping in a state visit that produced few concrete results on the technological front where the two powers collide.
The document urged Washington to lock in a lead of one to two years over China in artificial intelligence, arguing that Chinese labs have stayed near the frontier only by smuggling restricted American chips, renting them remotely in offshore data centres beyond the reach of US export law, and harvesting the output of American models to clone their capabilities.
It sketched two futures for 2028, one in which “democracies write the rules of the technology” and another in which Chinese firms spread cheap, capable AI across the developing world on the back of state subsidies and “good enough” hardware.
