Introduction: After Computing Power, Data Becomes the New "Normandy"
In August 2026, standing at the Q3 juncture and looking back at the AI frenzy of the past two years, a clear paradigm shift is evident. If 2024 was the year of the "computing power arms race" and 2025 was the year of "diverse model architectures," then 2026 is undoubtedly becoming the inaugural year of "data capitalization."
As financial reports from HK-listed cloud service providers reveal early signs of "computing power overcapacity," market focus on AI is shifting from underlying GPU stacking to upper layers. During this shift, a core contradiction is becoming prominent: as the marginal utility of large model parameter growth diminishes, high-quality, compliant, and rights-confirmed training data has become the biggest bottleneck constraining AI evolution. For investors, this means the market's vane has shifted from "shovel sellers" (computing power) to "miners" (data).
This "AI Intelligence Bureau" installment will delve into the "data rights" storm unfolding in Q3 2026, analyze how this trend reshapes the valuation logic of HK tech stocks, and reveal which sub-sectors are welcoming a golden window for value revaluation amidst the industry shift from "ocean fishing" to "paid angling."
Bidding Farewell to "Ocean Fishing": The End of the Free Data Era
In the early days of the ChatGPT explosion, the AI training model was vividly termed "ocean fishing." Tech giants indiscriminately scraped public data from the internet—from social media posts and e-books to news sites. While this aggressive growth approach initially brought leaps in model intelligence, it also buried huge legal and quality risks.
Entering 2026, this model is no longer sustainable. On one hand, regulatory scrutiny on copyright protection in major global economies has tightened unprecedentedly. From the full implementation of the EU's "AI Act" to the landing of multiple US court rulings on AI training data infringement, legal red lines have been clearly drawn. On the other hand, high-quality public data on the internet is about to be "eaten up." Industry estimates suggest that by the end of 2025, general large models had scrubbed almost all high-quality English text data; the remaining low-quality data not only fails to improve model performance but actually increases "model hallucinations."
Against this backdrop, relevant tech companies on the HK stock market are forced to adjust strategies. Startups still relying on crawler technology to acquire data face huge compliance costs and litigation risks. The market is realizing that data assets without a copyright moat are like castles built on sand, liable to collapse at any moment due to a single lawsuit.
Turning to "Paid Angling": The Commercial Closed Loop of Data Capitalization
When the free lunch ends, the AI industry is forced into the era of "paid angling." This is not as simple as buying datasets; it is a reconstruction of the industrial chain involving data production, rights confirmation, trading, and profit sharing.
The so-called "paid angling" refers to AI vendors having to establish formal commercial partnerships with holders of high-quality data, acquiring data through licensing, joint ventures, or revenue sharing. This transition has spawned two important market trends:
- Revaluation of Data Copyright Value: Companies possessing exclusive, high-quality, and long-term accumulated data suddenly hold the "oil" of the AI era. This includes publishers with decades of academic literature repositories, internet platforms with massive user behavior data, and vertical sector leaders holding specialized industry data.
- Upgrade of Synthetic Data and Data Annotation: To compensate for the shortage of real data, synthetic data technology saw an explosion in 2026. However, synthetic data is not created out of thin air; it still needs to be trained based on real-world logic and distributions, making the demand for "foundational data" even more urgent.
At this stage, we observe an interesting phenomenon in the HK stock market: internet giants with massive user ecosystems (like Tencent and Alibaba) demonstrate stronger risk resilience and commercialization potential than pure large model vendors due to their internal closed loops of high-quality data. Their "walled garden" strategy has turned into a solid moat in the era of data rights.
Policy Tailwind: The Deepening of China's "Data Element" Market
If global copyright litigation is the external pressure driving data rights, then China's strategic layout for "data elements" is the internal core driver. In 2026, a series of policies promoted by the National Data Bureau entered the deep water zone, and "data capitalization" (putting data assets on financial statements) has been fully popularized among central SOEs and large tech companies.
For the HK stock market, this is an investment logic with distinct Chinese characteristics. Unlike Western markets emphasizing personal privacy and copyright, the Chinese market emphasizes the public attributes and circulation value of data. Under this policy orientation, a batch of tech companies focusing on data trading, data cleaning, and data asset evaluation have begun to emerge.
Especially in key sectors like finance, healthcare, and industry, the circulation of data elements is being strictly regulated and accelerated. For example, in the medical AI field, medical informatization companies possessing desensitized clinical medical record data and imaging data are seeing their valuation logic shift from traditional software service providers to "AI data fuel suppliers." This leap in valuation is one of the important drivers behind the rebound of the HK stock biotech sector in Q3 2026.
Industrial Chain Value Reassessment: Finding HK Stock's "Data Water Sellers"
Based on the above analysis, we believe that under the "data rights" storm, the value revaluation of the HK stock AI industrial chain will mainly focus on the following three links, which are also directions investors need to focus on deploying:
1. Industry Leaders with Exclusive Vertical Data
The bonus period of general large models is over; future competition lies in deep applications in vertical industries. Enterprises that have plowed deep in specific industries (like financial investment research, biomedicine, and high-end manufacturing) for years and possess proprietary industry data will become the objects of competition for AI giants. Such companies can not only obtain high recurring revenue by licensing data but also use AI to empower their own businesses, achieving cost reduction and efficiency improvement.
2. Data Governance and Security Vendors in AI Infrastructure
As data becomes a core asset, data security, privacy computing, and compliance governance have become crucial. Cybersecurity and privacy computing targets on the HK stock market are shifting from pure "defensive spending" to "compliance necessities." Especially companies capable of providing cross-domain data circulation solutions (like federated learning technology) will occupy a core niche in the data element market.
3. "Data Asset Monetization" of Internet Platforms
For Chinese internet giants listed in HK, while advertising remains the foundation, data capitalization is becoming the second growth curve. By desensitizing data from scenarios like e-commerce, social networking, and entertainment to train vertical models, or opening API interfaces to third-party developers, these platforms are mining the stock value of data. The market's current valuation of this is still conservative, leaving a significant gap in expectations.
Independent Market Commentary: Beware of "Data Bubbles," Return to Business Essence
Although "data rights" brings huge imagination space, as independent market observers, we must also calmly point out potential risks. Currently in the HK stock market, some individual stocks touching on the "data element" concept have shown signs of over-hyping, with P/E ratios far exceeding what their fundamentals can support.
Investors must recognize that owning data does not equal owning assets. For data to become an asset, it must possess three characteristics: scarcity, compliance, and usability. Many companies claim to possess massive data, but if this data is low-quality, unstructured, or cannot prove ownership legally, then this data is not only worthless but may even become a liability (compliance cost).
In addition, the pricing mechanism for data assets is still in an exploratory stage globally. The lack of unified assessment standards leads to uneven quality of the "data assets" item in financial reports. When screening targets, we should focus more on companies whose data businesses have generated substantive cash flow and have clear practical cooperation with leading large model vendors, rather than those staying at the stage of concept hype.
Conclusion: "Cognitive Equality" and "Data Hegemony" in the Second Half of AI
In Q3 2026, the AI industry is experiencing a profound reshuffle. The popularization of computing power has achieved a certain degree of "cognitive equality" in algorithms, but the privatization and rights confirmation of high-quality data are forming a new "data hegemony."
For HK stock investors, this is both a challenge and an opportunity. We need to penetrate the fog of technology and see the return to business essence. In the second half of AI, algorithms may converge, computing power may be in surplus, and only high-quality, clearly rights-confirmed data is the non-replicable core asset. Those enterprises that have mastered "data hegemony" are very likely to enact a long-term bull market similar to that of internet platforms in the past decade in the future capital market. Let us stay sharp and grasp the pulse of this round of value revaluation.
