US allows Chinese telecom giant ZTE to purchase Nvidia H200 AI chips

The recent approval by the U.S. government for ZTE, a major Chinese telecom equipment manufacturer, to purchase NVIDIA's latest-generation H200 'Hopper' AI chips highlights the ongoing complexity of U... Read original →

The U.S. approval for ZTE to acquire NVIDIA’s H200 chips reveals a strategic inconsistency in its tech containment policy. Technically, while H200 is the final iteration of the Hopper architecture, its 192GB HBM3e memory significantly boosts AI training efficiency for Chinese cloud providers, delaying domestic GPGPU adoption. Compliance-wise, case-by-case licensing increases supply chain volatility, pushing Alibaba and Tencent toward costly multi-sourcing or pre-stocking strategies. Competitively, Huawei will leverage this to promote its Ascend 910B as a 'fully controllable' alternative, while NVIDIA races to monetize H20/H200 sales before Blackwell faces stricter bans. Over the next 12–24 months, Washington may expand selective licensing to ease pressure on U.S. chipmakers but simultaneously tighten controls on EUV and advanced packaging tools, trapping China’s AI infrastructure in a structural bottleneck defined by second-tier chips on lagging nodes.