DeepSeek V4 Will Run on Huawei Chips
DeepSeek’s upcoming V4 artificial intelligence model is set to run on Huawei Technologies’ latest chips. The move was reported by The Information, which said major Chinese technology companies, including Alibaba, ByteDance, and Tencent, have already placed bulk orders totaling hundreds of thousands of Huawei chip units ahead of the model’s expected release in the coming weeks.
This development follows months of preparation by DeepSeek to make sure its next flagship model launches as a showcase for Chinese-made hardware instead of the American chips that have long dominated AI development.
DeepSeek’s Hardware Choice
A deliberate move away from U.S. chipmakers
DeepSeek’s choice to prioritize domestic chipmakers did not happen overnight. The shift has been building since at least February, when it was reported that the Hangzhou-based lab had denied Nvidia and AMD early access to V4. That stands out because major chipmakers are usually given advance time to optimize their hardware for new models.
Instead, DeepSeek gave that advantage to Huawei and Cambricon Technologies, giving Chinese chip producers several weeks to fine-tune their processors for the new model.
Months of code work for domestic hardware
DeepSeek also spent months working with Huawei and Cambricon to rewrite portions of V4’s core code for their hardware. That effort points to a broader push within China’s AI industry toward hardware self-reliance.
The pressure behind that push has grown as U.S. export controls on advanced semiconductors have tightened. In that environment, adapting a flagship model to domestic chips carries technical value as well as strategic importance.
Huawei’s 950PR Chip Is Becoming More Popular
Bulk orders reflect rising confidence
The reported orders tied to DeepSeek V4 come at the same time industry interest is growing around Huawei’s newest AI processor, the 950PR. Customer testing has reportedly produced positive results, and ByteDance and Alibaba are among the companies planning orders.
That matters because the chip is being positioned for practical AI deployment, not just as a symbolic domestic alternative.
Built for inference workloads
The 950PR is designed to perform well on inference workloads. In simple terms, that means running trained AI models to answer queries, rather than handling the training stage, which requires much more raw compute power.
That distinction is important. If DeepSeek V4 is optimized for Huawei hardware in a real deployment setting, the emphasis is not just on building a model, but on serving and operating it efficiently.
Delivery outlook for Huawei’s new AI processor
Huawei is expected to deliver roughly 750,000 units of the 950PR this year. Mass production is expected to begin soon, with full shipments in the second half of 2026.
Those expected deliveries help explain why large Chinese technology companies are placing major orders ahead of V4’s launch. Hardware availability is a key part of whether a model rollout can scale.
DeepSeek V4 and China’s AI Self-Reliance Push
A flagship launch designed to showcase Chinese hardware
DeepSeek appears to be positioning V4 as more than a model release. The launch is being prepared as a demonstration that advanced AI systems can run on Chinese-made hardware.
That framing gives the model a larger role inside China’s technology landscape. It is not only about performance. It is also about proving that domestic chips can support a leading AI model at a meaningful scale.
Collaboration with Huawei and Cambricon
The decision to give Huawei and Cambricon a head start, while withholding early access from Nvidia and AMD, shows a clear preference for domestic ecosystem development. By aligning software optimization with local chipmakers, DeepSeek is helping create a tighter link between model design and Chinese semiconductor platforms.
That kind of coordination can strengthen the broader local AI stack, especially when a flagship model is involved.
The Geopolitical Meaning Behind the Shift
Political weight is hard to ignore
The move carries clear geopolitical undertones. A senior U.S. official previously said that DeepSeek’s model was trained using Nvidia’s advanced Blackwell chips, which are restricted for export to China.
Against that backdrop, DeepSeek’s decision to optimize V4 for Huawei hardware and exclude American chipmakers from early access has been widely viewed as a strategic effort to show that China’s AI ecosystem can operate independently of U.S. technology.
A long-term signal for the Chinese market
The broader interpretation is that DeepSeek’s approach may be part of a longer-term strategy to reduce Nvidia’s position in the Chinese AI market. The reported view is that DeepSeek will block Nvidia’s pre-model access and present V4 as being based on Chinese AI semiconductors such as Huawei’s.
Taken together, the decisions around access, optimization, and deployment point to more than a product launch. They suggest an attempt to shift how AI infrastructure is built and promoted inside China.
What the DeepSeek V4 Launch Suggests About the AI Market
Inference is becoming a key battleground
The focus on Huawei’s 950PR also highlights where competition is intensifying. The chip is aimed at inference, the stage where trained models are actually used to respond to prompts and requests.
That makes the V4 rollout especially notable. If the model runs effectively on Huawei chips in live use, it could strengthen confidence in domestic processors for real-world AI workloads.
Demand from major Chinese tech companies
Alibaba, ByteDance, and Tencent placing large orders adds commercial weight to the story. These are not minor buyers testing the waters. Their orders suggest serious demand ahead of the expected V4 release and reinforce the idea that the launch is tied to a broader ecosystem push.
When bulk orders and model optimization happen at the same time, it usually signals coordinated momentum rather than isolated experimentation.