JPM China AI

Asia Pacific Equity Research 16 August 2026

China Artificial Intelligence GLM-5.3 and DeepSeek's repricing shift the capability- cost frontier; raise Zhipu/MiniMax PT, remain OW/Neutral

Following Zhipu’s GLM-5.3 launch and DeepSeek’s API repricing effective Hong Kong Aug 17, we maintain OW on Zhipu and Neutral on MiniMax. We raise our Internet 2026-27 revenue forecasts by 6-9% and 0-11%, respectively, and lift our PTs Olivia Xu AC to HK$1,800 and HK$260. GLM-5.3 raises Zhipu’s own capability, while ( MiniMax benefits from a key rival becoming more expensive, and Hailuo H3 adds multimodal optionality. We prefer Zhipu as its improvement is internally SAC Registration Number: S driven and more defensible. For MiniMax, the upcoming M3.1 is the key test Alex Yao of whether its LLM can regain relevance on the capability-cost frontier. ( • Capability and cost define the Pareto frontier; we currently place SAC Registration Number: S

greater value on the intelligence end. A model sits on the frontier when no Daniel Chen rival offers better capability at the same or lower price, or comparable ( capability more cheaply. Both ends can support attractive SAC Registration Number: S commercialization: frontier intelligence can command premium pricing and J.P. Morgan Securities (China) Company unlock harder workloads, while cost-efficient models can capture large Limited volumes of mature intelligence. We currently favor capability because the intelligence frontier is still moving quickly, stronger models are relatively insulated from price changes by weaker alternatives, while mature intelligence is increasingly supplied by more providers. Sustainable cost leadership therefore requires structural efficiency, with DeepSeek providing the clearest example in China.

• GLM-5.3 moves Zhipu further up the capability curve; M3.1 is the next test for MiniMax. GLM-5.3 uses the same base model as GLM-5.2 but achieves meaningful coding and agentic gains through heavier post-training, according to the company. With API economics broadly unchanged, we expect stronger adoption and retention. MiniMax’s current M3 has yet to establish a clear advantage on either capability or cost-performance. DeepSeek’s repricing narrows one important cost disadvantage, but such benefit is externally driven. We therefore look to M3.1 to assess whether MiniMax can establish a more competitive position on either dimension of the Pareto curve.

• Raise PTs to HK$1,800/HK$260; H3 adds optionality while M3.1 is the key LLM catalyst for MiniMax. Our Zhipu PT implies ~20x our US$5bn 2027E ARR and MiniMax ~12x our US$1bn end-2026E ARR. H3’s encouraging early reception strengthens MiniMax’s multimodal portfolio, although we remain measured on the revenue and profit an independent provider can capture against integrated platforms including ByteDance and Kuaishou that can capture value across model/API revenue, content creation, distribution, advertising and user engagement. For Zhipu, we watch GLM- 5.3 validation and paid conversion; for MiniMax, M3.1 capability and pricing, H3 monetization and the sustainability of benefits from DeepSeek’s repricing are the key watchpoints.

See page 16 for analyst certification and important disclosures, including non-US analyst disclosures.

Olivia Xu Asia Pacific Equity Research ( August 2026

Equity Ratings and Price Targets Mkt Cap Price Rating Price Target Company Ticker ($ mn) CCY Price Cur Prev Cur End Prev End Date Date Z AI Co Ltd - H 2513 HK 39,020 HKD 1,270.00 OW n/c 1,800.00 Dec-26 1,600.00 n/c MiniMax Group Inc - H 100 HK 9,566 HKD 329.00 N n/c 260.00 Dec-26 160.00 n/c Source: Company data, Bloomberg Finance L.P., J.P. Morgan estimates. n/c = no change. All prices as of 14 Aug 26.

Capability and cost increasingly define the competitive frontier We remain constructive on the model layer as AI usage expands across coding, reasoning, agents and enterprise workflows, but we expect the economics to concentrate around models that maintain a clear advantage in either capability or cost-performance. We therefore increasingly assess model competition through a Pareto framework: a model sits on the frontier when peers cannot provide higher capability at…