Hong Kong-listed artificial intelligence (AI) concept stock Zhipu (02513.HK) surged strongly after Monday’s (July 13) market open, fueled by a dual boost from its founder’s internal letter and a foreign brokerage’s target price upgrade. Intraday gains exceeded 7%, with the stock reaching a high of HK$1,759 (approximately NT$7,200), on trading volume of HK$3.01 billion (approximately NT$12 billion), making it the market’s focal point for the day.
Market sources indicated that Zhipu founder Tang Jie sent an internal letter to all employees on July 11, officially announcing the launch of a new strategy called the “Touch High” initiative. In the letter, Tang explicitly stated that while the industry is broadly accelerating commercial monetization, Zhipu will adopt a “counterintuitive” approach, strategically deploying resources over the next two years without pursuing short-term application revenue, instead aiming directly for the next technological frontier of artificial general intelligence (AGI). He emphasized: “The great wave has arrived, and the trend is irreversible,” and “Failure to reach the summit is failure itself.”
Tang dissected his fundamental assessment of AGI competition in the letter, pointing out that genuine business opportunities never arise from fine-tuning products and models, but from leaps in the “upper bound of intelligence.” He believes AI is completing the leap from perceptual intelligence to cognitive intelligence, where machines no longer merely see and hear, but begin to understand and reason. To this end, the “Touch High” initiative will concentrate efforts on scaling four technological peaks.
The four core engines are: first, “long-range tasks,” enabling AI to evolve from real-time Q&A to handling complex capabilities for grand-scale engineering; second, “autonomous agent systems,” aimed at evolving from intelligent assistants to digital employees, building an agent society comprising thousands of agents with distinct professional personalities and skills, capable of autonomous debate, collaboration, code review, and resource scheduling; third, “fully self-supervised training,” converting computing power into evolutionary fuel by constructing high-quality synthetic data factories and AI self-play to achieve knowledge breakthroughs, while empowering the system to reconstruct its own code within a secure sandbox; and fourth, the element Tang emphasized most—”extreme security governance”—arguing that the more powerful the capability, the more robust the safety constraint mechanisms must be.
Beyond the long-term vision in the internal letter buoying market sentiment, JPMorgan’s consecutive target price upgrades also injected a shot of confidence for bulls. Following last Tuesday’s increase of Zhipu’s target price from HK$1,800 to HK$2,000, JPMorgan analyst Olivia Xu, in her latest report this week, again raised the target price by 20% to HK$2,400 (approximately NT$9,800), maintaining an “Overweight” rating.
Olivia Xu noted in the report that Zhipu’s demand is approaching the upper limit of its service capacity, and the company’s recently completed $4 billion (approximately NT$130 billion) share placement financing provides critical capital support for the already visible “demand-to-revenue” conversion pathway. She analyzed that the newly added inference computing resources are expected to convert into annual recurring revenue (ARR) within 12 months, making the commercialization outlook clearer. However, Olivia Xu also specifically highlighted the risks, noting that since Zhipu’s free float post-placement is only 14%, volatility will remain extremely high for a stock that has already surged 14-fold this year.
Notably, JPMorgan offered a different assessment of another AI company, MiniMax, in the same report. The bank lowered its target price for MiniMax from HK$300 to HK$240 (approximately NT$1,000), maintaining a “Neutral” rating. The report argued that MiniMax’s commercial conversion pathway remains unproven and that its equity has been diluted. JPMorgan indicated that its stance would only turn more positive upon observing clear evidence of model improvement and signs of narrowing the monetization gap with peers.
Additionally, Zhipu has recently made new technical advances. Tang Jie simultaneously announced in his internal letter that the company has released its most capable open-source model to date, GLM-5.2. The model supports genuinely usable million-token context, maintains leadership in long-range tasks, and is open to all users. More strategically, Zhipu decided to officially open-source GLM-5.2 under the most permissive MIT license, meaning anyone can download, deploy, and commercialize the model without entity-based restrictions. Market analysts believe this move not only demonstrates Zhipu’s technical confidence but also helps rapidly expand its developer ecosystem and industry influence.
Driven by the dual catalysts of the “Touch High” initiative’s long-term layout and JPMorgan’s consecutive bullish reports, Zhipu’s stock price demonstrated strong resilience against selling pressure. Institutional investors noted that while the market holds divergent views on AI concept stock valuations, Zhipu’s choice to bet counter-cyclically on fundamental technological breakthroughs amid an industry monetization wave could, if it successfully conquers challenges like long-range tasks and autonomous agents, create distance from competitors in the next phase of the AGI race.