DeepSeek confirms Huawei-based V4 inference: "After the 950 supernodes are launched at scale in the second half of this year, the price of Pro is expected to be reduced significantly."
DeepSeek's V4 inference on Huawei chips promises massive price cuts by H2 2025.
DeepSeek has officially confirmed that its V4 inference model will be deployed on Huawei-based hardware, marking a strategic shift toward domestic Chinese semiconductor infrastructure. The announcement, made via a company representative, reveals plans to launch 950 supernodes at scale in the second half of 2025. These supernodes—high-performance computing clusters optimized for AI workloads—are expected to dramatically increase inference capacity while reducing operational costs.
"After the 950 supernodes are launched at scale in the second half of this year, the price of Pro is expected to be reduced significantly," the company stated. This price reduction could make DeepSeek's Pro tier more competitive against global AI providers like OpenAI and Anthropic. The move also signals growing confidence in Huawei's Ascend AI chips, which are designed to compete with Nvidia's GPUs. For enterprises, this could mean more affordable access to high-performance AI inference, particularly for applications requiring large-scale deployment. However, the timeline depends on successful scaling and potential geopolitical hurdles affecting chip supply chains.
- DeepSeek V4 inference will run on Huawei hardware with 950 supernodes planned for H2 2025
- Significant price reductions expected for Pro tier after scaling
- Strategic pivot to domestic Chinese chips reduces reliance on Nvidia GPUs
Why It Matters
Huawei-based AI inference could lower costs and boost China's tech independence.