JPMorgan says competitive AI models that publish open-source weights broaden
adoption and monetise via official APIs, partnerships, enterprise deployments
and workflow products, while weaker models face faster price comparison and
traffic diversion. JPMorgan raised Zhipu’s 2026–30 revenue forecasts by 3–9%,
narrowed adjusted net losses for 2026 and 2027 to RMB3.71bn and RMB3.14bn
respectively, and revised a prior 2028 adjusted loss of RMB1.29bn to an adjusted
profit of RMB2.37bn; PT raised to HK$2,000, rating maintained Overweight.
JPMorgan flags GLM‑5.5/6, KimiK3 and DeepSeekV4.1 performance as critical to
Zhipu’s ability to sustain a lead. For MINIMAX‑W (00100.HK), JPMorgan cut
2027–30 revenue forecasts by 2–8%, lowered PT to HK$300 and kept Neutral, noting
the M3 model carries a permanent 50% discount that signals no clear capability
premium versus domestic peers; JPMorgan says the view would turn positive if
MiniMax narrows the capability gap, the discount normalises, API volumes hold
and MiniMaxCode increases workflow stickiness.
