Key Takeaways:
JPMorgan raised its December 2026 price targets for Zhipu to HK$1,800 and MiniMax to HK$260, citing GLM-5.3 and DeepSeek's API price increases.
"Models with frontier-intelligence pricing power hold more investment value than those relying purely on low prices," JPMorgan said in an Aug. 16 research note.
The targets are based on 20 times 2030 estimated earnings, discounted at a 15 percent weighted average cost of capital to December 2026. Zhipu's 2026-30 revenue forecasts were raised 6-10 percent, while MiniMax's 2027-30 forecasts rose 11-21 percent.
The upgrades come as DeepSeek raises API prices for V4 Pro and V4 Flash effective Aug. 17, tripling peak-hour input costs to 9 yuan per million tokens. The hike narrows the cost gap for rivals and strengthens the case for capability-led competition.
JPMorgan kept Zhipu at Overweight, arguing the GLM-5.3 upgrade — built on the same base model as GLM-5.2 with gains from reinforcement post-training — is internally driven and builds a more durable moat. The bank said data quality, reinforcement learning, evaluation infrastructure and engineering execution now carry more weight in differentiating competitors, without a new round of large-scale pretraining. With GLM-5.2 pricing unchanged at 8 yuan input, 2 yuan cached input and 28 yuan output per million tokens, stronger capability should lift adoption and retention.
MiniMax was kept at Neutral. JPMorgan said the M3 model has not established a clear edge in capability or cost-performance, leaving it squeezed between stronger models such as Kimi K3 and GLM-5.3 on one side and DeepSeek's aggressive pricing on the other. The bank called M3.1 the company's most important upcoming release, with Hailuo H3 adding multimodal option value, though it flagged that integrated platforms such as ByteDance and Kuaishou can capture value across more segments.
On DeepSeek, JPMorgan said the price increase shows the company retains pricing flexibility and remains the industry's cost benchmark. The hike offers near-term breathing room for cost-focused rivals, but sustainable cost leadership still requires structural efficiency advantages such as DeepSeek's MoE architecture and KV-cache optimization.
The target increases imply 12.5 percent upside for Zhipu and 62.5 percent for MiniMax from prior levels, reflecting JPMorgan's view that capability-led models will command premium valuations. Investors will watch MiniMax's M3.1 launch and Zhipu's next model iteration for confirmation of the thesis.
This article is for informational purposes only and does not constitute investment advice.
