Chinese artificial-intelligence start-up DeepSeek has postponed the release of its next-generation R2 model after repeated attempts to train the system on Huawei’s Ascend processors failed, according to the Financial Times. Engineers working alongside Huawei staff were unable to achieve a stable training process on the Chinese chips, forcing DeepSeek to switch back to Nvidia H20 hardware for model training while continuing to plan deployment on Huawei chips for inference. The R2 model had been slated for launch in May but now has no confirmed release date, pushing availability beyond the end of August. The setback underscores the technical gap that still separates China’s domestic semiconductors from U.S. alternatives despite export controls, and it highlights the difficulties companies face as regulators press them to replace foreign technology in strategic sectors.