Morgan Stanley raises MiniMax (00100) target price, Chinese model leverages global token market at 1/5 cost.

On April 28th, Morgan Stanley Asia-Pacific research team released a new industry report titled "China's AI Path: More Bang For The Buck," significantly raising the target price of two leading AI companies in China. They are optimistic about the "cost-effectiveness narrative" of China's AI path continuing to pay off in terms of ARR and valuation. In particular, the target price of MiniMax (00100) was raised from 990 Hong Kong dollars to 1100 Hong Kong dollars, with the industry rating maintained as "In-Line," and the focus on recommended assets being companies with full-stack AI capabilities such as MiniMax. The gap between China and the United States in AI is narrowing to 3-6 months, with engineering efficiency becoming China's competitive advantage. The report points out that the Scaling Law still holds, with the cutting-edge LLM capabilities in the United States breaking through faster than expected, but Chinese AI companies have not hit a wall in terms of engineering efficiency "scaling." With computing power still posing a hard constraint, the top models in China and the United States are almost equal in terms of the Artificial Analysis Intelligence Index - MiniMax M2.7, KNOWLEDGE ATLAS GLM-5.1, Moonshot K2.6, DeepSeek V4 all fall within the 50-54 point range. Morgan Stanley predicts that the gap between China and the United States in AI has narrowed to 3-6 months. The cost-effectiveness advantage of Chinese models is mainly reflected in achieving the same level of intelligence as their U.S. counterparts at 15%-20% of the inference cost. The report attributes the improvement in engineering efficiency to three main directions: dense models and MoE at the architectural level, attention mechanism improvements; reinforcement learning and model distillation at the post-training level; and hardware optimization and KV cache efficiency at the inference infrastructure level. The recently released DeepSeek-V4 is seen as the latest example. There has been a historic turning point in token consumption share, with leading models gaining pricing power. Data from OpenRouter shows that the token consumption share of Chinese leading models rose rapidly from 5% in April 2025 to 32% in March 2026, while the U.S. leading models' share dropped significantly from 58% to 19%. MiniMax, KNOWLEDGE ATLAS, and Alibaba's token usage in February-March 2026 increased 4-6 times compared to December last year, mainly driven by breakthroughs in coding and agentic Workflow capabilities. Contrary to market concerns about "models being commoditized," cutting-edge models are showing clear pricing power: the price of KNOWLEDGE ATLAS has doubled since the beginning of the year, and MiniMax's KV cache price has also doubled. Morgan Stanley clearly expects that MiniMax will initiate a significant price increase after the upgrade to the M3 model, which may serve as a catalyst for the next stage of ARR growth. Expectation of 5x ARR growth in three years with Hong Kong liquidity structure facing reshaping. In terms of ARR forecasts, Morgan Stanley provides a baseline-optimistic range: the ARR of a single leading model is expected to reach $1-1.5 billion by the end of 2026, further climbing to $2.5-5 billion by the end of 2027, corresponding to an annual growth rate of 3-5 times. As of February 2026, MiniMax's ARR has reached $150 million, and KNOWLEDGE ATLAS's ARR in March has reached $250 million. On the Hong Kong Stock Exchange side, KNOWLEDGE ATLAS and MiniMax are highly likely to be included in the Hang Seng TECH Index on June 8th, with a combined weight of 5%-7%, expected to bring in passive funds of $1.25-1.75 billion. The report also points out that the technology sector accounts for 40% of IPO fundraising on the Hong Kong Stock Exchange from the beginning of the year until now, with 43% in reserve projects, and AI is becoming a long-term force reshaping the composition of Hong Kong stock indices, liquidity, and fund flow.