The narrative around AI in the capital markets is undergoing a subtle yet profound transformation.
On July 22, 2026 (UTC+8), Hong Kong’s AI large model sector faced significant selling pressure. MiniMax-W (HKG: 0100) closed at HKD 197.00, dropping 11.42% in a single day. Zhichu (HKG: 2513) closed at HKD 1,175.00, down 3.61%. Meanwhile, the Hang Seng Tech Index fell about 1.8%, with the overall AI concept sector underperforming the broader market.
This trend isn’t an isolated event. Since Q2 2026, the Hong Kong AI sector has seen similar corrections multiple times. The market is left wondering: Is this simply profit-taking after previous surges, or is there a structural shift underway in how AI valuations are determined?
This article analyzes the short-term market performance, explores the evolving global AI competitive landscape and shifting valuation frameworks for large model companies, and examines the stress tests facing MiniMax and Zhichu. We’ll also discuss where the next phase of AI investment logic may lead.
Short-Term Pullback: Profit-Taking and Valuation Adjustment
From a trading perspective, this pullback was triggered by clear short-term factors.
Since Q1 2026, MiniMax’s cumulative gains exceeded 80%, while Zhichu’s rose over 45% in the same period. With no near-term performance catalysts, some investors opted to lock in profits. The July 22 sell-off was accompanied by a surge in trading volume—MiniMax’s turnover was roughly twice the average of the previous five trading days, indicating concentrated selling.
Global risk appetite contracted in tandem. As the Federal Reserve’s July policy meeting approached, uncertainty around the pace of rate cuts increased. The US dollar index strengthened, and emerging market tech sectors came under broad pressure. Hong Kong, as an offshore market, is especially sensitive to global liquidity expectations. High-beta AI concept stocks often bear the brunt of these shifts.
These factors explain the short-term volatility, but they don’t account for the clear divergence within the sector—MiniMax’s decline far outpaced Zhichu and the broader market. This suggests the sell-off wasn’t indiscriminate; instead, it reflected strong stock-specific factors or differentiated pricing.
A deeper question emerges: How much of current AI stock prices reflect a premium for "future technological leadership," and how much is tied to "current commercial monetization"? Over the past two years, AI large model company valuations have been driven largely by narratives around technical capability and parameter scale. As the story moves toward realization, investors are demanding more traditional financial metrics—revenue, customers, profit margins. This shift inevitably loosens the old valuation logic.
Competitive Landscape: From "Tech Race" to "Multi-Dimensional Warfare"
Competition in the AI large model sector has clearly intensified.
For the past two years, the main focus has been "who has more model parameters" and "who ranks higher on benchmarking lists." Technical breakthroughs alone fueled waves of revaluation. But as we enter 2026, this single-dimensional competition is giving way to a multi-dimensional commercial battle.
Globally, OpenAI, Google Gemini, Anthropic, and Meta AI form the first tier. Their common traits: parent companies or financing capabilities in the hundreds of billions, ecosystem access to hundreds of millions or even billions of users, and massive self-built compute clusters. These elements create a "flywheel effect" for model iteration—more users generate more feedback, more feedback improves the model, and improved models attract more users.
In China, competition is equally fierce. Alibaba Tongyi Qianwen leverages Alibaba Cloud’s enterprise network and e-commerce scenarios; Tencent Hunyuan integrates into WeChat and gaming ecosystems; Baidu Wenxin benefits from search entry points and years of AI development; Byte Doubao rapidly scales with Douyin’s traffic advantage. The common thread among these tech giants: model capability is not their only competitive moat—sometimes, it’s not even the most important one.
For independent large model companies like MiniMax and Zhichu, the challenge has shifted from "can you build a good model" to "can you establish a sustainable commercial loop amid the giants."
Valuation Logic Shift: Markets Are Redefining "Moats"
The valuation framework for large model companies is transitioning through three stages.
The first stage is driven by technical leadership. Capital markets are willing to pay high premiums for parameter scale, benchmarking scores, and published papers. Here, a company’s value is determined by the "potential possibilities" of its technology.
The second stage is driven by commercialization validation. The market starts to focus on metrics like API call growth, enterprise paying customers, changes in average revenue per user, and revenue structure. Investors now care more about whether "technical advantages translate into revenue advantages."
The third stage is driven by ecosystem and cost. As foundational model capabilities converge, competition centers on two points: First, whoever has lower unit inference costs can take the lead in pricing, expanding their customer base and forming a data flywheel. Second, whoever has more application scenarios and user touchpoints can generate a positive cycle of data and traffic.
Currently, the market is at a critical juncture between the first and second stages. Companies that consistently disclose commercialization progress and demonstrate clear revenue paths will see their valuations supported. Those still stuck in "technical narrative" mode will face ongoing valuation pressure.
Judging by the different declines of MiniMax and Zhichu, the market may already be applying this screening logic—investors are pricing in differentiated commercialization prospects for each company.
Breaking Down Competitive Pressures for Independent AI Vendors
MiniMax’s core strengths lie in its multimodal capabilities and early overseas market expansion. But its challenges are equally specific: On the consumer side, it must compete for user engagement with products like Byte Doubao and Baidu Wenxin, which have massive traffic entry points. In the enterprise market, it faces bundled "model + cloud service" solutions from cloud providers like Alibaba and Tencent. On the cost front, independent vendors are naturally disadvantaged in GPU procurement scale and bargaining power, directly impacting API pricing competitiveness.
Zhichu’s technical accumulation and open-source ecosystem are key moats, especially giving it a first-mover advantage in enterprise services. Its main pressures come from competing with internet giants’ cloud ecosystems—enterprise clients often prefer AI solutions tied to their existing cloud providers. Commercialization speed is another challenge—while big players can spread AI costs across multiple business lines, independent vendors must recoup every R&D dollar from their AI business. As foundational model capabilities converge, establishing differentiated technical branding becomes crucial.
Both companies share a common challenge: With sustained high investment in foundational models, can revenue growth cover costs—or at least show a clear path toward coverage?
Conclusion
The recent pullback in Hong Kong’s AI large model stocks is more likely the start of a structural valuation reset than the end of the AI trend.
As competition intensifies among global and Chinese AI giants like OpenAI, Google, Alibaba, and Tencent, the core competitiveness of large model companies is shifting from single "technical leadership" to a multi-dimensional contest of "model performance + cost efficiency + commercialization capability + ecosystem scale."
For investors, this means the window for AI investment is narrowing—from "sector beta" to "stock alpha." Future winners won’t be all model companies, but those that can prove sustainable revenue growth, optimize cost structures, and build enduring customer relationships.
AI technology continues to evolve rapidly, but capital market patience is thinning. As the tide recedes, the difference between those merely swimming naked and those truly going the distance will become clearer than ever.
FAQ
1. What are the main reasons behind the declines in MiniMax and Zhichu stock prices?
In the short term, after substantial gains in AI concept stocks, there’s pressure for profit-taking, compounded by a drop in global risk appetite due to uncertainty around Federal Reserve policy. The deeper reason is a shift in market valuation logic—from "technical leadership narratives" to "commercialization validation." Investors are scrutinizing each company’s revenue growth, customer acquisition, and cost control. Companies lacking a clear path to profitability face valuation discounts.
2. Who are the main competitors for Chinese AI large model companies?
Internationally, the main competitors are OpenAI, Google Gemini, Anthropic, and Meta AI—all backed by massive capital and global user ecosystems. In China, competition is equally intense, with key rivals including Alibaba Tongyi Qianwen, Tencent Hunyuan, Baidu Wenxin, and Byte Doubao. These internet giants not only have strong R&D capabilities but also rich application scenarios and huge user bases, enabling ecosystem synergy beyond the models themselves.
3. How has the valuation logic for large model companies changed?
Previously, valuations focused on "potential indicators" like model parameter scale and technical rankings. Now, the market is demanding quantifiable commercialization outcomes—API call revenue, enterprise paying customers, and realized AI application revenue. At the same time, compute cost control (including GPU investment and inference costs) and ecosystem breadth are becoming key valuation factors. Investors are prioritizing sustainable profitability over pure technical vision.
4. Does the adjustment in Hong Kong’s AI sector signal the end of AI investment opportunities?
It doesn’t mark the end of the AI trend, but rather a shift in investment logic from "sector-wide opportunities" to "selective stock picking." Going forward, the market will reward companies with clear revenue growth paths, cost advantages, and ecosystem moats, while those lacking fundamental support may remain under pressure. For investors, the opportunity window is moving from betting on the whole sector to carefully choosing companies with lasting competitiveness.
5. What are the core commercialization challenges facing AI large model companies?
The primary challenge is whether high R&D and compute costs can be covered by revenue growth—including GPU cluster investment, data training costs, and inference expenses. Next is customer acquisition and retention, as independent AI vendors must compete with internet giants that already have established enterprise relationships. As foundational model capabilities converge, companies need to find differentiated application scenarios and industry solutions to avoid pure price competition. Loss of API pricing power and lack of ecosystem are additional risks independent vendors must watch out for.
