DeepSeek V4 Set to Run on Huawei Chips as China Doubles Down on Homegrown AI Infrastructure
China’s artificial intelligence landscape is shifting dramatically as DeepSeek’s upcoming V4 model is confirmed to be built to run on Huawei Technologies’ latest chips, according to a report published by The Information on Friday. This marks a significant milestone in China’s push to develop a fully independent AI technology stack, free from reliance on American semiconductor giants.
The development signals a deliberate and strategic move by Chinese AI developers to pivot away from US chip manufacturers like Nvidia, especially amid escalating trade restrictions that have severely limited China’s access to high-performance Western processors.
In anticipation of V4’s imminent launch, some of China’s largest tech conglomerates have already made substantial hardware commitments. Alibaba Group, ByteDance, and Tencent Holdings have collectively placed bulk orders for Huawei’s next-generation chips, totalling hundreds of thousands of units. Five individuals with direct knowledge of the purchases confirmed the details to The Information.
According to the report, the DeepSeek V4 model is expected to launch within the next few weeks, though no official date has been confirmed. Neither Huawei nor DeepSeek responded to media requests for comment, as queries were submitted outside of regular office hours.
DeepSeek Collaborates With Huawei and Cambricon to Rewrite Core Code
One of the most notable aspects of this development is the depth of collaboration between DeepSeek and domestic chip designers. Over the past several months, DeepSeek engineers have been working side by side with both Huawei and Cambricon Technologies, another prominent Chinese semiconductor company, to rewrite key sections of V4’s underlying code. Two people close to the company confirmed this collaboration during the testing phase.
This isn’t just a matter of porting an existing model onto new hardware. The level of code-level modification required suggests that DeepSeek is engineering V4 from the ground up to perform optimally on Chinese silicon — a technically demanding process that reflects the seriousness of China’s domestic AI ambitions.
Beyond a single flagship release, DeepSeek is reportedly developing two additional V4 variants, each tailored for different capabilities and designed specifically to run efficiently on Chinese-made chips. This multi-variant strategy further underscores how comprehensively the company is aligning its roadmap with domestically produced hardware.
Earlier this year, Reuters reported that DeepSeek had notably broken from standard industry practice by not sharing its upcoming flagship model with US chipmakers for performance optimisation — a step most AI labs typically take before major model updates. Instead, DeepSeek granted early access exclusively to domestic suppliers, with Huawei being among the first to receive it.
This deliberate exclusion of American companies from the optimisation process sends a clear message about the direction Chinese AI development is heading — one that is increasingly self-reliant and strategically independent from Western tech ecosystems.
DeepSeek is no stranger to market-shaking moments. Earlier this year, the release of its low-cost models V3 and R1 triggered a significant selloff across global technology stocks. Investors began questioning whether US AI firms truly needed to pour billions into expensive computing infrastructure, given that a Chinese lab had apparently achieved comparable results at a fraction of the cost.
That disruption has only intensified global curiosity about what DeepSeek V4 will bring once it officially arrives. With major Chinese tech players betting heavily on Huawei hardware and DeepSeek engineering its models specifically for domestic chips, the race to build a fully independent Chinese AI stack appears to be well underway — and moving faster than many in the West had anticipated.