China’s ByteDance, Alibaba and Tencent move to lock in $35B of Nvidia AI chips
Chinese tech giants ByteDance, Alibaba and Tencent have reportedly ordered more than 3 million Nvidia RTX Pro 5500 GPUs worth over $35 billion, highlighting the rush to secure hardware needed to train and run powerful AI models under tightening export controls.
A small group of Chinese technology giants is moving to secure a huge share of advanced AI hardware, underscoring how graphics chips have become a strategic resource.
ByteDance, Alibaba and Tencent have together ordered more than 3 million Nvidia RTX Pro 5500 graphics processing units (GPUs) worth over $35 billion, according to people cited by international media on 20 October. These chips are used to train and operate large artificial intelligence models for recommendation systems, language processing, image generation and video.
If completed on these terms, the order would be one of the largest known single commitments for AI‑class GPUs. It would lock in years of supply for China’s internet and cloud leaders at a time when the U.S. is tightening export controls on high‑end chips for Chinese customers and when demand from American cloud providers and startups is already stretching Nvidia’s production capacity.
For engineers at ByteDance, Alibaba and Tencent, guaranteed access to this hardware determines how quickly they can build and deploy new models. Training a frontier‑scale language model can occupy tens of thousands of GPUs for weeks. Owning millions gives these firms the ability to train several large systems in parallel and to sell AI services into sectors such as finance, gaming, e‑commerce and advertising.
The scale of the order will be felt by competitors. Smaller Chinese AI labs and cloud companies that already struggle to obtain top‑tier Nvidia hardware may find themselves waiting even longer. Outside China, other buyers for the RTX Pro 5500 line will have to compete with a customer capable of committing tens of billions of dollars in a single deal, depending on how Nvidia allocates its capacity under export rules.
Governments are watching for national‑security reasons as well as economic ones. U.S. and European officials treat advanced AI chips as dual‑use technology with applications in areas such as military simulation, surveillance and information operations. A long‑term commitment by Chinese platforms to several million of these units raises questions about how current export controls are working and whether further restrictions or licensing conditions might follow.
On the Chinese side, the reported order shows how important it is for Beijing’s tech sector to cushion itself against future sanctions and supply shocks. With a large stockpile of compatible Nvidia chips, major platforms can keep running their existing architectures while they try to develop domestic accelerators and software that is less dependent on foreign hardware.
Building and powering the data centers required for 3 million GPUs will also reshape local infrastructure. Operators will need much larger facilities, steadier high‑voltage power supplies, industrial‑scale cooling and stronger security. Policymakers are already debating how to protect these AI and cloud centers as critical infrastructure rather than ordinary server halls.
Investors and officials will be looking for detail from Nvidia on its order backlog and on any export‑compliant variants for China, as well as for any response from Washington on chip‑control lists. Signals from Beijing on domestic GPU development and from Europe on its own compute plans will show how other players react to the scale of China’s latest move.
Sources
- OSINT