China is preparing to permit some artificial intelligence companies to acquire Nvidia's H200 graphics processing unit, a chip designed for AI workloads. The move marks a notable departure from Beijing's earlier strategy of restricting purchases to bolster domestic tech independence.
For months, the United States and China were locked in a stand-off over advanced semiconductors. In early 2026, Washington approved the H200 for export to China, but Beijing deliberately blocked Chinese firms from buying them as part of its self-sufficiency push.
The H200 is a cutting-edge GPU that powers AI model training and inference. The shift suggests Beijing may be prioritizing immediate AI capabilities over its long-term goal of reducing reliance on foreign technology.
Selected companies will be allowed to purchase the chip, though specific firms and timelines remain unspecified. The policy change could accelerate AI development in China while easing tensions with Washington.
The reversal may also pressure domestic chipmakers, who have been racing to produce alternatives. It remains unclear how broad the approvals will be or what conditions might apply.