China’s leading AI lab DeepSeek will run its next-generation model V4 on chips designed by Huawei Technologies, U.S. outlet The Information reported on Friday, a move that signals deeper alignment between Chinese AI developers and domestic semiconductor suppliers. The report, cited by Reuters, said major Chinese internet groups including Alibaba, ByteDance and Tencent have placed bulk orders for Huawei’s upcoming chips — totaling “hundreds of thousands” of units — ahead of V4’s expected launch in the coming weeks.
According to people briefed on the deals, DeepSeek has spent recent months working directly with Huawei and fellow Chinese chip designer Cambricon Technologies to adapt and rewrite parts of V4’s underlying code so the model runs efficiently on local silicon. Two sources close to DeepSeek told The Information that the lab is also building two additional V4 variants, each optimized for different capabilities and explicitly targeted to run on Chinese-made chips.
The arrangement departs from standard industry practice, The Information noted, because DeepSeek has not granted U.S. chipmakers early access to its flagship model for performance optimisation. Instead the lab has prioritised domestic suppliers, a choice Reuters previously reported. Neither Huawei nor DeepSeek responded to requests for comment outside normal office hours, the report added.
The flurry of chip orders from Alibaba, ByteDance and Tencent underscores the commercial stakes for V4. DeepSeek’s earlier low-cost releases, V3 and R1, prompted a global tech stock selloff last year by intensifying questions about how much U.S. firms needed to invest in AI computing infrastructure. V4 — long-awaited by investors and industry watchers — has attracted significant attention because it could reshape demand for AI accelerators and reshape competitive dynamics in the global chip market.
DeepSeek’s decision to co-develop optimisations with Huawei and Cambricon also reflects a broader push in China to build an AI stack that is less dependent on Western hardware. Domestic chipmakers have been racing to close performance gaps with foreign rivals while courting large cloud and internet customers for volume business. For major platforms such as Alibaba and Tencent, securing chips compatible with leading local models is increasingly vital for deploying services at scale.
If V4 launches on schedule in the coming weeks and the reported chip orders translate into wide deployment, the episode could accelerate the commercialisation of Chinese AI hardware and deepen ties between national champions in cloud, chips and AI research. Observers will be watching not only the model’s capabilities but also whether the technology and supply-chain choices alter competitive dynamics between Chinese and U.S. companies in the fast-moving AI ecosystem.