Morgan Stanley's Asia-Pacific research team has released a new industry report, increasing the target price for MiniMax from HKD 990 to HKD 1100. The firm maintained its "In-Line" industry rating, highlighting MiniMax and other companies with full-stack AI model capabilities as key recommendations.

The report states that, while computing power remains a key constraint, top Chinese and US AI models have reached near parity on the Artificial Analysis Intelligence Index. Models such as MiniMax's M2.7, Zhipu's GLM-5.1, Moonshot's K2.6, and DeepSeek's V4 all scored within the 50-54 point range. Based on this data, Morgan Stanley concludes that the AI gap between the US and China has narrowed to just 3-6 months.

Morgan Stanley highlighted the cost-effectiveness of Chinese models, noting they achieve comparable intelligence levels at only 15-20% of the inference costs of their US counterparts. The report attributes these engineering efficiency gains to three main areas: architectural improvements involving dense models and Mixture-of-Experts (MoE), along with enhanced attention mechanisms; post-training refinements such as reinforcement learning and model distillation; and inference infrastructure optimizations including hardware enhancements and improved KV cache efficiency.

According to OpenRouter data, the market share of token consumption for leading Chinese models surged from 5% in April 2025 to 32% by March 2026. During the same period, the share for leading US models dropped significantly from 58% to 19%. Token usage for MiniMax, Zhipu, and Alibaba increased by 4 to 6 times between February and March 2026 compared to December of the previous year, primarily driven by breakthroughs in coding and agentic workflow capabilities. Morgan Stanley anticipates that MiniMax will implement significant price increases following the rollout of its M3 model upgrade.