DeepSeek Orders 160,000 Huawei Ascend 950DT AI Chips for 1GW Data Center Facility

Frontier AI laboratory DeepSeek has finalized plans to deploy at least 160,000 of Huawei's next-generation Ascend 950DT AI accelerators at its massive new 1-gigawatt data center in Ulanqab, Inner Mongolia. The procurement represents one of the largest single enterprise silicon commitments outside the NVIDIA ecosystem, designed to transition DeepSeek's high-volume user inference workloads—including DeepSeek-V4-Pro—to domestic Chinese semiconductor hardware.

Scaling Inference on Huawei Ascend 950DT

Backed by over $7 billion raised in June 2026 to transition from leased cloud compute to proprietary physical facilities, DeepSeek's Ulanqab site will utilize the Ascend 950DT to serve global chatbot and coding agent API traffic. While DeepSeek continues utilizing existing NVIDIA clusters for primary model pre-training, the 950DT chips offer dedicated INT4/FP8 matrix engines and high-bandwidth interconnects optimized specifically for token generation at scale.

Deployment Architecture & Specifications

• Deployment Scale: At least 160,000 Ascend 950DT accelerator modules scheduled across initial phases.
• Power Capacity: 1-gigawatt high-density computing campus situated in Ulanqab, Inner Mongolia.
• Workload Specialization: Dedicated entirely to production inference, API routing, and agentic reasoning for DeepSeek-V4-Pro.
• Supply Chain Resilience: Bypasses western export licensing restrictions while supporting China's domestic silicon self-reliance.

Strategic Shift in Global AI Compute

The massive purchase establishes Huawei as a viable hyperscale inference provider capable of fulfilling multi-hundred-thousand-unit datacenter contracts, validating alternative semiconductor supply chains as western foundries face mounting packaging and memory shortages.

Procurement Summary