Nvidia's US manufacturing plans: Announced to manufacture AI chips and build supercomputers in the US for the first time, commissioning over one million square feet of manufacturing space across Arizona and Texas.
- Comes amid US-China tensions and Trump administration's push for domestic manufacturing.
- Less than two weeks after Trump administration's chaotic tariff rollout and two days after contradictory messages on electronic component exemptions.
- Has already begun producing Blackwell chips at TSMC's Phoenix facility; setting up supercomputer manufacturing plants in Texas with partners like Foxconn and Wistron.
- Supply chain details: For chip packaging and testing, collaborating with Amkor and SPIL in Arizona. TSMC's Phoenix facility is used for front-end processing, but advanced packaging might still need to be done in Taiwan.
- H20 chip situation: Avoided export controls by striking a domestic manufacturing deal with the Trump administration. Can still be exported to China with reduced specifications while maintaining compatibility with CUDA platform.
- Economic and job aspects: Aims to produce up to half a trillion dollars of AI infrastructure in four years, creating "hundreds of thousands" of jobs and driving "trillions of dollars" in economic activity. But there are uncertainties due to trade policy and lack of skilled workers. Trump has pressured TSMC directly.
- Automation in US manufacturing: Plans to use its own technologies like NVIDIA Omniverse and NVIDIA Isaac GR00T in US facilities to create digital twins and build robots for automation.
**粗体** _斜体_ [链接](http://example.com) `代码` - 列表 > 引用
。你还可以使用@
来通知其他用户。