China is allowing limited shipments of Nvidia's H200 chips to enter the mainland, with ByteDance and Tencent each receiving around 10,000 processors.
While Chinese AI companies have released competitive models, they lag far behind US providers in inference capacity — the computing power needed to serve users.
This allocation helps narrow that gap while respecting US export controls that prohibit shipments of Nvidia's most advanced chips to China.
What happened
China is allowing small shipments of Nvidia's H200 processors into the mainland to support domestic AI companies. ByteDance and Tencent each received around 10,000 H200 chips, with more companies potentially receiving allocations soon. The H200 is at least two generations behind Nvidia's most powerful chips, which are blocked from China by US export controls.
Why it matters
Chinese AI labs have released competitive models (DeepSeek V4, Alibaba's Qwen3.8, Moonshot's K3, and Z.ai's GLM-5.3) but lack the inference capacity — the computing power needed to serve users at scale — that US providers command. When demand surged, Moonshot had to turn customers away, signaling a real bottleneck. The H200 allocation lets domestic firms narrow this gap while staying within US export rules, which permit up to 100,000 H200s per company.
What to watch
Purchases require sign-off from China's NDRC planning agency, and shipments to Hong Kong are also allowed, though the city lacks sufficient data centers and power to absorb large volumes. Nvidia holds around half a million H200s in stock. Beijing is signaling it will continue backing domestic chip makers like Huawei in parallel.
Ask the AI about this article →
China's decision to permit H200 shipments reflects a pragmatic accommodation within US export restrictions. Rather than ban all advanced Nvidia chips to China, US policy allows the H200 — an older generation — up to 100,000 units per company. China is leveraging this window to equip its leading AI firms while avoiding direct breach of sanctions. The move is targeted: purchases require approval from China's NDRC planning agency, and Beijing is simultaneously backing domestic alternatives like Huawei.
The underlying constraint is not model quality but inference capacity. Chinese AI labs have demonstrated technical parity or near-parity with US models: Alibaba's Qwen3.8, DeepSeek's V4, Moonshot's K3, and Z.ai's GLM-5.3 represent genuine competitive offerings. However, inference — the real-time compute needed to serve millions of users — requires massive hardware infrastructure. Moonshot's inability to accommodate surging customer demand despite a technically strong model illustrates the gap. By allocating H200s to ByteDance and Tencent, China is addressing this capacity shortfall directly, buying time while domestic chip development and manufacturing (led by Huawei) matures.
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytime
Ask AI anything about this article. Q&As are published on this page for other readers too.
Anthropic is privately hoping to file for its initial public offering by the end of this month, targeting a ra…
Broadcom is reportedly seeking to borrow up to $100 billion in debt financing to support growth efforts at Ant…
Elice Group, a South Korean AI infrastructure provider, announced the launch of the country's first AI data ce…
On August 12, AT&T's Chief Data and AI Officer said OpenAI models power about 25% of the telecom's total AI us…
On August 11, IBM announced a multi-year $240 million agreement with Together AI to deploy NVIDIA HGX B300 sys…
Tencent Holdings is receiving shipments of Nvidia H200 AI chips under a new Chinese policy that permits limite…
