On 20 October 2026, Chinese technology companies ByteDance, Alibaba and Tencent had collectively ordered more than 3 million RTX PRO 5500 graphics cards from Nvidia, with a stated combined value exceeding 35 billion. The cards cost about 11,400 each and are designed for professional workstations.
Graphics Processing Units and AI Inference
A graphics processing unit (GPU) is a specialised processor that performs many calculations in parallel. The RTX PRO 5500 is a workstation GPU, rather than a data-centre processor designed for large, linked computing clusters. The Chinese companies intend to use the cards mainly for AI inference: running a trained artificial-intelligence model to produce outputs such as chatbot responses. Inference generally requires less computing power than training a model on large datasets.
Nvidia’s B300 is a Blackwell-based data-centre chip, while its H200 uses the Hopper architecture. The B300 is barred from sale to Chinese buyers under United States export controls; H200 exports are subject to case-by-case review and licensing. The RTX PRO 5500 is among Nvidia’s more powerful products still available to Chinese buyers under the restrictions.
Export Controls and Chip Supply
Nvidia has said it cannot guarantee delivery of the orders because demand from Chinese technology companies exceeds planned production. One estimate given to partners was that they might receive 5 to 10 per cent of the quantities requested. Nvidia has indicated that global production of RTX PRO 5500 cards could be limited to no more than 3 million in 2027, while orders from Chinese customers have exceeded 6 million.
Export controls are government restrictions on the export of specified goods or technologies to particular destinations or users. The United States Commerce Department’s Bureau of Industry and Security oversees US export controls. The supply constraints concern a professional workstation GPU, while the B300 and H200 are data-centre chips.
Production Capacity and Semiconductor Manufacturing
Taiwan Semiconductor Manufacturing Co. (TSMC) is expected to receive a significant portion of Nvidia’s production capacity for its upcoming Rubin data-centre platform from 2027. The reported Rubin production shift coincides with the projected global production limit for the RTX PRO 5500.
Important Facts for Exams
:
- AI inference is the process of running a trained model to generate an output.
- The RTX PRO 5500 is a workstation graphics card, not a data-centre chip.
- The Bureau of Industry and Security oversees United States export controls.
- TSMC is the abbreviation for Taiwan Semiconductor Manufacturing Co.
Nvidia’s B300 uses the Blackwell architecture, and its H200 uses the Hopper architecture.
