Nvidia is tuning its hardware to run China's top open-source AI models better, a defensive move against Huawei that now risks running into new US restrictions.

• Nvidia is making its chips the best place to run Chinese open models like DeepSeek, Qwen and Kimi.
• The goal is to keep China's huge developer community on American technology instead of Huawei's Ascend chips.
• In its latest quarterly filing, Nvidia warned that new US limits on Chinese AI models could hit its core business.
• Washington lawmakers are uneasy, and there is talk of fresh chip tariffs or an outright ban.
• Microsoft, Meta, Palantir and more than 20 other US tech firms urged policymakers in July not to clamp down on open models too early.

Outlook: Nvidia's China strategy hinges on whether the White House moves ahead with restrictions on Chinese-built AI models.