🎙 Listen to a summary of this story
China’s biggest technology companies have reportedly placed orders worth more than $35 billion for a new Nvidia graphics processor, a purchasing surge that exposes both the scale of the country’s artificial intelligence ambitions and its continuing dependence on American-designed chips.
ByteDance, Alibaba and Tencent ordered more than 3 million RTX PRO 5500 graphics cards, according to an Oct. 20 report by Reuters, which cited two people familiar with the matter. The cards carry a reported price of about $11,400 each.
The companies did not respond to Reuters’ requests for comment, and Nvidia has not confirmed the individual orders, their combined value or whether it can deliver the requested volumes. The figures should therefore be treated as source-based reporting rather than confirmed purchasing disclosures.
Even so, the reported orders offer a striking measure of demand for computing power inside China. They suggest that the country’s leading technology groups are prepared to build AI systems around less powerful workstation processors because access to Nvidia’s most advanced data-centre products remains restricted.
A workstation chip becomes an AI alternative
The RTX PRO 5500 was designed principally for high-end professional computers rather than the vast clusters normally used to train frontier AI models. Nvidia’s official product documentation describes it as a Blackwell-architecture processor with 84 gigabytes of GDDR7 memory and support for generative AI, autonomous software agents, simulation and computer-vision workloads.
Nvidia lists the product as “coming soon,” meaning the reported orders concern hardware that is not yet commercially available. Reuters said customers expect samples in December and broader deliveries during the first quarter of 2027.
Chinese buyers reportedly plan to use the cards mainly for inference — the process of running trained models to generate answers, images or other outputs. That generally requires less computing power than training a large model from the beginning, making workstation cards a possible substitute for restricted data-centre processors.
They are not a direct replacement for Nvidia’s H200 or newer B300 chips. Those products are designed to operate in large, tightly connected installations capable of handling demanding training and deployment workloads. But the RTX PRO 5500’s substantial memory and AI-focused components could allow Chinese companies to expand services such as chatbots, recommendation systems and automated content generation.
That distinction matters for ByteDance, Alibaba and Tencent. All three operate digital platforms serving hundreds of millions of users, meaning the expense of running AI models after training can become as important as the cost of developing them.
Export controls reshape the market
The buying rush is unfolding inside a complicated system of U.S. restrictions. Washington has sought to limit China’s access to processors and manufacturing technology that could support advanced AI and military capabilities.
Rules are not identical across every product. In January, the U.S. Commerce Department’s Bureau of Industry and Security said applications to export Nvidia H200 chips and comparable processors to China would be reviewed individually, subject to security and compliance requirements. More advanced hardware remains unavailable or tightly controlled.
Nvidia has repeatedly warned that restrictions can cost American companies sales while encouraging Chinese competitors. Its latest quarterly filing identifies export controls, customer concentration and access to manufacturing capacity among significant business risks.
In response to the Reuters report, Nvidia said Chinese workstation-chip suppliers had experienced rapid growth since the first U.S. controls were imposed in 2022. The company argued that American suppliers faced pressure from both U.S. restrictions and Chinese limits on imported products.
Beijing has encouraged technology companies to use domestically designed processors. Huawei has emerged as the strongest Chinese challenger in AI hardware, while other semiconductor developers are expanding production. Yet the size of the reported Nvidia orders indicates that domestic supply, software compatibility or performance may not yet satisfy all of China’s demand.
That creates a mixed outcome for Washington. Restrictions appear to be denying Chinese companies unrestricted access to the most capable American processors and stimulating investment in domestic alternatives. At the same time, Chinese groups may be adapting by buying large numbers of compliant chips and combining them for workloads they can still perform.
A $35 billion order that may not be filled
The largest uncertainty is supply. Reuters reported that total Chinese demand for the RTX PRO 5500 exceeded 6 million cards, while Nvidia may produce no more than 3 million worldwide during 2027. Some customers were told they could receive only 5% to 10% of the quantities requested.
Nvidia must also allocate manufacturing capacity among competing products. The company is preparing its Rubin platform for next-generation data centres, potentially giving higher priority to more advanced and profitable processors. Nvidia disclosed $279 billion in supply-and-capacity commitments in its latest quarterly filing, illustrating the extraordinary scale at which it is securing components and production.
There is some public evidence of Chinese demand beyond the three reported buyers. Reuters said Shenzhen-listed Tansun Technology disclosed an agreement to acquire 20,000 RTX PRO 5500 cards, while warning that supply constraints and technology controls could reduce or delay deliveries.
But the headline value of the ByteDance, Alibaba and Tencent orders is based on reported list prices. It may not represent money already paid or the eventual cost after commercial negotiations. Orders can also be reduced, delayed or cancelled, particularly when regulatory approvals and scarce production capacity are involved.
The deeper significance lies less in the precise $35 billion figure than in the behaviour it describes. China’s technology leaders appear unwilling to pause their AI expansion while waiting for domestic processors to mature or U.S. policy to change. They are instead seeking enormous quantities of the strongest Nvidia hardware they can still obtain — even when that hardware was not originally intended to anchor China’s next generation of AI infrastructure.
