DeepSeek V4 and Chinese Tech Giants Shift to Huawei AI Chips

DeepSeek is preparing to launch its V4 artificial intelligence model using chips designed by Huawei Technologies, according to reporting from The Information. The shift marks a decisive move toward domestic semiconductor independence for the Hangzhou-based AI startup, reducing its reliance on hardware from U.S. Chipmakers.

In preparation for the V4 launch, several Chinese technology giants—including Alibaba Group, ByteDance, and Tencent Holdings—have placed bulk orders for Huawei’s upcoming chips totaling hundreds of thousands of units. The next-generation model is expected to be released within a few weeks of April 3, 2026.

Domestic Hardware Optimization

To ensure the V4 model operates efficiently on domestically produced hardware, DeepSeek spent several months collaborating with Huawei and Cambricon Technologies. This partnership involved rewriting portions of the model’s underlying code during the testing phase.

In a departure from standard industry norms, DeepSeek reportedly denied early access to Nvidia and AMD for performance optimization of its upcoming flagship model. Instead, the lab granted this early access to domestic suppliers.

DeepSeek is also developing two additional V4 variants. These versions are being optimized for different capabilities and are built specifically to run on Chinese chips.

Strategic and Geopolitical Drivers

The transition to Huawei hardware is driven by a combination of supply constraints, the need for strategic control over AI infrastructure, and geopolitical pressures. U.S. Export controls, backed by Donald Trump, have restricted China’s access to cutting-edge hardware, such as Nvidia’s Blackwell architecture.

By optimizing its models for Huawei chips, DeepSeek is attempting to future-proof its operations against further U.S. Restrictions. Huawei’s Ascend chips are intended to offer lower costs and scalable inference performance.

Despite this shift, Nvidia continues to lead globally in high-end AI training power. Early reports indicate that hybrid hardware usage may continue to maintain system stability.

Market Context

The rollout of V4 follows the release of DeepSeek’s low-cost V3 and R1 models. Those releases triggered a global tech stock selloff last year, leading investors to question whether U.S. AI firms required billions of dollars in spending on AI computing power.