[News] DeepSeek V5 Could Complete Training on Huawei Ascend 950 in 2027, Moving Beyond Inference

What happened

According to Caiwen News citing industry sources, DeepSeek plans to train its next-generation V5 model primarily on Huawei's Ascend 950 chips, with the project expected to materialize in 2027; the report stresses this could happen, and V5 is still in the selection phase with model specifications, key capabilities, and launch timeline undisclosed, while industry sources expect a release sometime in 2027 depending on training progress. Per Star Market Daily, Huawei's Ascend chips have so far been used primarily for inference rather than training, and V5 could become DeepSeek's first large language model trained natively on Ascend 950 chips. The source states the cause: with U.S. export restrictions limiting access to NVIDIA's latest high-end AI chips, previously approved H-series GPUs are reportedly struggling to meet the training demands of next-generation models such as DeepSeek V4 and V5, making domestic alternatives increasingly critical; a January Reuters report noted DeepSeek-V3 required just 2.788 million H800 GPU-hours for full training. Per Caiwen, Ascend remains DeepSeek's primary training platform with no reported adoption of solutions from other Chinese chipmakers including Moore Threads, and Huawei holds no stake in DeepSeek, with cooperation limited to products, computing resources, and technical optimization. On Huawei's side, per Star Market Daily citing Rotating Chairman David Wang, the Ascend 960DT is now expected to be ready in 1Q27, three quarters ahead of its original schedule; per Guancha, the Ascend 960PR is slated for 3Q27, one quarter ahead of schedule, followed by the Ascend 970 in 2028 and the Ascend 980 in 2029.

Impact and signal

Affected components and suppliers

Source and time

Source: TrendForce · 11/2/2026 · View original source