In China, H3C, one of the largest server manufacturers, informed its customers in a statement about a possible shortage of the Nvidia H20 chip, the most advanced artificial intelligence processor legally available in the country under US export controls. This is reported by UNN with reference to Reuters.
Details
The potential supply crisis could create obstacles for China's artificial intelligence ambitions at a time when its technology firms are actively expanding their investments in artificial intelligence.
The international supply chain of H20 faces significant uncertainty.
H3C also noted in the statement that it would allocate incoming H20 chips on a "profit first" basis, prioritizing stable, long-term customers with higher profit margins on their orders.
H20 processors are currently in short supply in the Chinese market, said a source familiar with the situation, who distributes AI servers, including those using H20 chips.
We were told the chips would be available, but when it came time to buy them, we were told they had already been sold at higher prices.
In January, Reuters reported that US officials were considering restricting sales of H20 chips to China.
H3C is a key OEM partner for Nvidia's AI chips in China, alongside other companies such as Inspur, Lenovo, and Huawei's x86 server xFusion.
Nvidia's main chip legally allowed to be sold in China, the H20, was released after the latest round of US export restrictions took effect in October 2023.
The US previously banned the export of Nvidia's most advanced chips to China since 2022. This was due to concerns that its competitor could use advanced technologies to build up its military capabilities. Analysts estimate that Nvidia shipped about one million H20 units in 2024, generating more than $12 billion in revenue for the company.
Addition
Chinese fintech giant Ant Group, backed by Alibaba founder Jack Ma, has used Chinese-made semiconductors to develop methods for training artificial intelligence models that could cut costs by 20%.
Earlier this year, DeepSeek briefly crashed Nvidia's stock on speculation that its models require far fewer chips.