On November 18, during its Q3 2026 earnings call, Tencent addressed the impact of computing power restrictions on its AI business. President Martin Lau stated that both leased and self-built computing power are facing constraints, which inevitably affects model training. As a result, the launch of the next version of the Hunyuan model may take longer in the short term. He noted that computational efficiency in model training and inference is improving rapidly year over year, and that the company's previous Hy4 training approach had already achieved relatively high efficiency. Over the long run, the domestic supply chain will gradually be integrated into the model inference stage. At present, the domestic industry is developing rapidly, and the localization process is expected to accelerate further.
