Nvidia CEO dismisses AI chip diversion claims

Nvidia chief executive Jensen Huang stopped in Taipei during Computex 2025 to insist that “there’s no evidence of any AI chip diversion” to restricted markets. He told Bloomberg that data-center systems built around the new Grace Blackwell architecture weigh close to two tons and ship as integrated racks, making covert exports impractical.

Huang also argued that customers and governments know the rules and “monitor themselves very carefully” because they want to keep buying Nvidia hardware. He described Washington’s decision to scrap earlier “AI diffusion” controls as the right move, echoing his long-held view that limiting U.S. technology overseas is “precisely wrong.”

The remarks come as the Middle East—particularly the United Arab Emirates and Saudi Arabia—seeks more Nvidia accelerators to fuel domestic AI projects. Huang said proper production planning could satisfy that demand without reshuffling existing allocations.

Critics point out that reality is messier. Reports this year detail Chinese buyers obtaining H-class GPUs through shell companies in Malaysia, Vietnam, and Taiwan; one reseller even flaunted contraband H200 boards on social media. Singapore has begun investigations, and U.S. officials have asked Malaysian authorities to curb what they call a fast-growing gray market after imports of advanced GPUs there jumped more than 3,400 percent.

Lawmakers in Washington are considering requiring manufacturers to enable geo-tracking on top-tier gaming and AI processors, arguing that weight alone is no barrier to smuggling—stolen automobiles, they note, cross borders regularly. Nvidia counters that once servers leave its hands, tracing individual boards is nearly impossible, highlighting the gap between regulatory aspirations and technical realities.