Kunlun Tech Led a Split Institutional Trade as Demingli Faced Heavy Selling

Kunlun Tech drew RMB 458 million in reported net institutional buying on July 31, while Demingli recorded RMB 430 million in net selling. The figures, distributed through the RSSHub 36Kr news feed and attributed to First Financial, point to a sharp split within Chinese technology trading.

Institutions appeared in the exchange disclosures of 42 stocks that day. Twenty-eight showed net institutional buying, while 14 showed net selling. BlueFocus ranked second among reported purchases at RMB 321 million, followed by GCL Energy Technology at RMB 148 million.

The selling side was concentrated in different parts of the technology market. Demingli led with RMB 430 million in reported net outflows. Semiconductor equipment supplier Hline Technology and electronic-materials company Yoke Technology followed at RMB 150 million and RMB 128 million.

Those numbers look like a vote on six companies. They are better understood as a snapshot of competing trades across artificial intelligence, digital advertising, energy infrastructure, memory hardware, semiconductor equipment, and chip materials.

The distinction matters because China’s Dragon and Tiger List does not reveal every trade or identify every institution. It discloses selected trading seats when a stock meets specified market-movement conditions. A large net purchase can show where institutions acted during an unusual session, but it does not establish a durable investment position.

What the July 31 Institutional Numbers Actually Show

The clearest signal was not broad confidence in technology stocks. It was selective demand for particular stories inside a volatile market.

According to the news item carried by the RSSHub 36Kr feed, institutional seats appeared in 42 stocks on July 31. The 28-to-14 division favored net buying, but the distribution remained uneven.

Kunlun Tech led the buying group with RMB 458 million in net institutional purchases. BlueFocus followed at RMB 321 million, while GCL Energy Technology recorded RMB 148 million.

These figures suggest that the strongest disclosed demand clustered around companies associated with artificial intelligence applications, digital marketing, and energy services. That grouping is notable, but it does not prove one coordinated sector trade.

The three companies expose institutions to different operating risks. Kunlun Tech is associated with internet services and generative AI. BlueFocus operates in marketing and communications, where AI can affect content production and campaign workflows. GCL Energy Technology belongs to the energy sector and carries a different set of demand, financing, and policy variables.

The selling list was similarly diverse. Demingli develops storage-related products. Hline Technology supplies semiconductor testing equipment and related systems. Yoke Technology sells electronic materials used in areas that include semiconductor manufacturing.

Their common connection is hardware exposure, yet their economics are not interchangeable. Memory products, testing equipment, and specialty materials sit at different points in the supply chain. Demand cycles, inventories, customer concentration, and capital requirements can diverge substantially.

The July 31 data therefore describes a divided market. Institutions bought some AI-linked and infrastructure narratives while selling selected semiconductor and storage names.

It is also important to distinguish reported institutional activity from total market flow. “Net institutional buying” usually reflects purchases minus sales attributed to disclosed institutional seats. It does not measure every fund, insurer, asset manager, or proprietary trading desk active in the stock.

A seat can also trade in both directions during one session. The net number compresses those transactions into a single result, hiding the gross buying and selling behind it.

The headline amounts remain useful because they identify where disclosed activity was concentrated. They become misleading only when treated as complete ownership data or an institutional consensus.

Why the Dragon and Tiger List Is a Narrow Market Window

The list reveals who traded through leading disclosed seats, not why they traded or how long they plan to stay.

China’s exchanges publish trading details for stocks that meet defined conditions. These disclosures are commonly called the Dragon and Tiger List in English-language market coverage.

For Shenzhen main-board stocks with daily price limits, the exchange can publish the five largest buying and selling members or trading units when specified thresholds are reached. The conditions include certain price deviations, trading ranges, and turnover rates, according to the exchange’s disclosure rules.

This means the list is event-driven. It does not provide a standardized daily ownership report for every listed company.

A company can appear because its shares moved sharply, traded through an unusually wide range, or changed hands at a high rate. That selection process creates an important bias: the disclosed stocks are already unusual.

The list then shows the largest qualifying seats within that unusual session. It does not identify all market participants, and an “institution-only” seat does not necessarily reveal the beneficial owner behind each order.

That makes the data different from a quarterly fund filing. A quarterly filing can show a reported holding at a point in time. The Dragon and Tiger List shows selected trading activity associated with a day of elevated market behavior.

The time horizon is another limitation. A disclosed buyer might be establishing a long-term position, covering a short exposure, executing an arbitrage strategy, or trading around an existing holding. The public table cannot reliably separate those motives.

The same problem affects net selling. An institution can reduce risk after a rapid gain without turning bearish on a company’s long-term business. It can also sell one account while another institutional account buys.

Investors should therefore read the July 31 figures as market microstructure data. Market microstructure describes how orders, trading rules, and participant behavior combine to produce transactions and prices.

That perspective changes the central question. The useful question is not whether “institutions like Kunlun Tech.” It is why institutional seats became unusually active in Kunlun Tech during that session.

The RSSHub 36Kr item provides the amounts and rankings, but not a verified account of each trader’s motive. No public seat table can supply that missing intent by itself.

This does not make the data worthless. It makes the evidence narrower than the headline suggests.

Kunlun Tech and BlueFocus Won the AI Narrative Trade

The buying leaders gave investors liquid exposure to AI applications, but one session cannot validate their underlying economics.

Kunlun Tech ranked first among the reported institutional purchases. Its position at the top directs attention toward the application side of China’s AI market rather than only the semiconductor supply chain.

That distinction is central to the July 31 split. Application companies can benefit from expectations of higher usage, new products, or improving monetization without carrying the same inventory risks as hardware vendors.

They still face substantial costs. Model training and inference require computing resources, while consumer applications must attract and retain users. Revenue can lag spending when companies prioritize product development or market share.

A buyer of Kunlun Tech is therefore balancing at least two ideas. The first is that generative AI can support new products and engagement. The second is that the resulting revenue will eventually justify development and computing costs.

The disclosed trading data confirms neither proposition. It shows that institutional buying exceeded institutional selling within the seats included in that day’s public information.

BlueFocus adds a different application-layer exposure. Marketing companies can use generative AI to produce variations of copy, images, and campaign concepts. They can also automate parts of media planning, customer analysis, and reporting.

The commercial question is whether those efficiencies improve margins or mainly reduce the price clients will pay. If competitors gain access to similar systems, productivity improvements can spread across the industry without producing a lasting advantage.

That tension helps explain why the July 31 buying should not be described as a simple endorsement of AI. Institutions can trade an expected earnings change, a product catalyst, or a short-term revaluation without making a broad claim about the technology.

GCL Energy Technology complicates the picture further. Its presence among the three largest net purchases prevents the list from being reduced to an AI software story.

Energy infrastructure increasingly matters to data-intensive computing, but the reported trading data does not establish that AI demand caused GCL’s institutional inflow. Energy markets also respond to power demand, project economics, policy, financing conditions, and asset utilization.

The more defensible interpretation is selective rotation. Investors favored three companies with distinct narratives while reducing exposure elsewhere.

The RSSHub 36Kr report gives a concise ranking. A serious analysis must preserve the separation between a reported flow and the business thesis that traders might have used.

That separation is especially important for international readers. Company names translated into broad labels such as “AI stock” or “chip stock” can conceal major differences in revenue sources, balance sheets, and customer exposure.

A practical reading starts with each company’s exchange filings. Shenzhen’s information rules require listed companies to disclose material information fairly and prohibit selective release of undisclosed material facts. The exchange’s disclosure guidance reinforces the line between public corporate information and market speculation.

Institutional trading cannot replace that evidence. It can tell readers where to investigate first.

Demingli Selling Challenged a Simple Technology Rally

Demingli’s RMB 430 million net outflow is the clearest counterweight to the buying headline because it shows institutions reducing another technology exposure simultaneously.

Demingli led reported institutional selling on July 31. Its RMB 430 million net outflow nearly matched the RMB 458 million net inflow recorded for Kunlun Tech.

That symmetry creates the day’s central tension. Institutions were not simply entering technology stocks. They were reallocating among different types of technology risk.

Demingli’s connection to storage products places it closer to the semiconductor inventory cycle. Storage markets can move rapidly when supply, customer inventories, device demand, and product prices change.

A rising product price can support revenue and margins, but it can also produce inventory gains that prove temporary. Falling prices can pressure suppliers and distributors even when unit shipments remain healthy.

The Dragon and Tiger List cannot show whether July 31 sellers were responding to valuation, inventory expectations, a company-specific development, or portfolio risk limits. Any claim about motive would go beyond the verified evidence.

Recent trading history also warns against treating “institutional” as one stable bloc. Separate market coverage showed institutional seats as net buyers of Demingli on an earlier July session, even while the stock also faced substantial total net selling. That contrast illustrates how rapidly seat activity can change.

A disclosed institution can buy during one volatile session and sell during another. Different institutions can also take opposing positions on the same day.

Hline Technology and Yoke Technology reinforce the hardware-side pressure, but their inclusion should not be interpreted as a unified rejection of semiconductors.

Hline serves parts of the semiconductor testing and equipment market. Capital-equipment demand depends on customer spending plans, factory utilization, product qualification, and the timing of expansion projects.

Yoke supplies specialty materials. Materials businesses face another combination of factors, including product mix, customer certification, input costs, and production capacity.

Selling across these three names can reflect a sector rotation. It can also result from unrelated company-level decisions that happen to appear together in one ranking.

The skeptical view is therefore straightforward. The July 31 flows are real reported transactions, but the narrative attached to them remains uncertain.

There is no verified statement from the institutions explaining why they bought Kunlun Tech or sold Demingli. There is also no evidence in the summarized report that the trades came from the same institutions.

This uncertainty weakens any claim that institutions collectively moved from hardware into software. It does not eliminate the possibility, but it means subsequent data must test the interpretation.

Investors should look for repetition. If application-oriented companies continue attracting institutional seats while hardware companies repeatedly face net selling, the rotation thesis becomes stronger.

If the pattern reverses within several sessions, July 31 will look more like event-driven trading than a durable capital shift.

What RSSHub 36Kr Readers Should Not Infer From One Session

The report identifies an important trading imbalance, but it cannot establish conviction, future returns, or superior information.

The first mistaken inference is that 28 net purchases imply a broadly bullish institutional market. The list included only stocks that generated reportable public trading information.

The full market contained many more securities. Without a comparable measure for all stocks, the 28-to-14 ratio cannot describe institutional sentiment toward the entire A-share market.

The second mistake is equating net buying with new ownership. An institution can add to an existing holding, repurchase shares after an earlier sale, or conduct a short-duration strategy.

The public data does not reveal the institution’s starting position. A net buyer that day might still have reduced its overall exposure across a longer period.

The third mistake is treating an institutional seat as a named investor. Exchange disclosures often use generic institutional classifications. They do not necessarily tell readers which fund made the decision or whether multiple accounts contributed to the total.

The fourth mistake is assuming institutional activity predicts the next price move. Large traders can be early, wrong, hedged, or constrained by mandates that do not apply to individual investors.

Price-limit systems also shape behavior. Research into Chinese price-limit events has examined whether limits cool trading or attract further orders near the boundary. One study found evidence of a cooling effect alongside a tendency toward next-day continuation in its historical sample, but that result cannot predict any single 2026 stock.

More importantly, historical market studies do not explain the motives behind Kunlun Tech’s July 31 buyers. They provide context for unusual trading conditions, not a company-specific forecast.

The fifth mistake is attaching every technology trade to artificial intelligence. Kunlun Tech and BlueFocus have clear AI-related narratives, but their disclosed inflows can reflect valuation, momentum, corporate developments, or other factors.

Likewise, selling in Demingli, Hline, and Yoke does not prove that institutions rejected China’s semiconductor strategy. It only shows net selling among disclosed seats for those stocks that day.

Readers should also separate the publisher from the underlying market record. The RSSHub 36Kr feed surfaced a report attributed to First Financial. The exchange disclosures remain the primary records for the individual trading entries.

RSSHub is an aggregation and delivery layer. It can make updates easier to monitor, but it does not add verification to the underlying claim.

This distinction matters in automated news systems. A repeated item can appear more important when several feeds reproduce it, even if every copy traces back to one report.

Researchers building a daily briefing should preserve the source chain. A feed entry should link back to the publisher, while key numbers should be checked against exchange data whenever practical.

A searchable knowledge base can help analysts retain those source relationships. The important feature is provenance, not simply collecting more headlines.

For this event, provenance means recording that the summarized figures were reported through a 36Kr newsflash, attributed to First Financial, and based on after-hours trading disclosures. It also means documenting what the report does not establish.

The unknowns include the identities of the beneficial owners, their starting positions, their strategy horizons, and the catalysts behind their orders.

Those gaps should remain visible. Filling them with confident speculation would turn a useful market signal into an unreliable investment story.

The Real Contest Is Narrative Momentum Versus Business Evidence

The July 31 split matters because institutional flows can amplify a market narrative before financial results confirm it.

Kunlun Tech and BlueFocus offer investors recognizable application-layer stories. Their products and services allow the market to discuss AI adoption without waiting for every part of the domestic hardware stack to mature.

That accessibility can attract capital during periods when investors favor revenue-facing applications. It can also increase sensitivity to product announcements, usage claims, and changes in market sentiment.

Hardware and materials companies sit under different expectations. Investors frequently evaluate them through inventory levels, capacity utilization, customer orders, product prices, and capital spending.

These metrics can produce a gap between long-term strategic importance and short-term earnings performance. A component can be essential to an industry while its supplier still faces pricing pressure or excess inventory.

The July 31 trading pattern placed those narratives on opposite sides of the disclosed institutional ledger. Application names led buying, while selected storage and semiconductor names led selling.

However, the opposition is not permanent. AI applications depend on chips, memory, networking, power, and data-center capacity. Strong application demand can eventually support hardware demand.

The timing can still differ. Software usage can grow before infrastructure spending reaches a supplier, while hardware orders can rise before applications generate sustainable revenue.

This makes earnings evidence the critical test. For Kunlun Tech, investors need to examine whether AI-related products contribute measurable revenue, user engagement, or improving unit economics.

Unit economics describes the revenue and direct cost associated with serving a user or transaction. It matters because an application can grow usage while losing more money on each additional interaction.

For BlueFocus, the central question is whether AI-supported production improves margins and client retention. Faster content generation alone does not guarantee higher profit if clients negotiate lower fees.

For GCL Energy Technology, investors need evidence that project performance and energy demand support cash generation. A broad data-center narrative cannot substitute for operating results.

Demingli requires a different checklist. Product prices, inventory levels, customer demand, and gross margins can reveal whether institutional selling anticipated weaker conditions or merely followed market volatility.

Hline’s orders and customer acceptance cycles offer another signal. Equipment revenue can shift between reporting periods when installation or qualification timing changes.

Yoke’s product mix and semiconductor-material demand will show whether the July selling matched a fundamental slowdown. Without those confirmations, the flow remains an incomplete clue.

This is why the July 31 event is more than an ordinary ranking. It captures the market deciding which evidence it expects to arrive first.

Application buyers appear willing to pay for nearer narrative momentum. Hardware sellers appear less willing to absorb uncertainty in selected cyclical or capital-intensive names.

That judgment can change quickly. A strong order update, improved margin, new customer qualification, or better product pricing can restore interest in the hardware group.

Weak monetization data can reverse the application trade just as easily. Institutional activity does not exempt any company from financial verification.

Three Signals Will Test Whether the Rotation Lasts

The next several months should reveal whether July 31 began a sustained reallocation or captured only a volatile trading day.

The first signal is repeated institutional activity in subsequent exchange disclosures. One day establishes a data point. Several aligned sessions can establish a pattern.

Watch whether Kunlun Tech and BlueFocus continue to attract net institutional buying when they next qualify for public trading disclosures. Repeated buying would strengthen the application-layer rotation thesis.

The reverse matters just as much. If Demingli, Hline, or Yoke soon record substantial institutional buying, the July 31 divide will look less durable.

Readers should compare gross purchases and sales when available, not only the net result. A large net number created by exceptionally high activity can describe a contested stock rather than unanimous conviction.

The second signal is company-reported operating evidence. Product usage, revenue mix, gross margins, inventories, and orders offer more durable information than seat rankings.

For the application companies, the strongest evidence would connect AI-related activity to revenue or improved efficiency. Management descriptions without measurable financial impact should receive less weight.

For storage and semiconductor companies, inventory and margin trends can show whether selling anticipated weaker fundamentals. Stable or improving results would weaken a bearish interpretation of the July 31 outflows.

All such evidence should come from public filings and investor communications available to the entire market. Shenzhen’s 2026 trading framework took effect on July 6 under the exchange’s revised trading rules, making current rules the relevant baseline for interpreting late-July activity.

The third signal is the relationship between prices and disclosed flows. A stock that keeps rising while institutional seats sell might be receiving demand from other participants. A stock that falls despite institutional buying shows that those seats did not control the overall market balance.

This relationship helps prevent a common analytical error. Institutional net buying is not the same as total net demand because every completed trade contains both a buyer and seller.

The disclosed categories organize selected orders. They do not change the basic structure of the market.

Investors should also watch whether volatility declines after the reportable session. A calmer market with stable prices suggests that the unusual trading was absorbed. Continued extreme turnover or price movement signals that positioning remains unsettled.

Together, these three tests cover persistence, fundamentals, and market response. They are more informative than trying to guess the identity of an unnamed institutional buyer.

The original RSSHub 36Kr item is best used as an alert. It tells readers where an unusual concentration of disclosed activity occurred and which companies deserve closer examination.

It should not function as a buy list. Nor should Demingli’s position at the top of the selling ranking become a stand-alone bearish thesis.

The July 31 disclosures presented a genuine conflict. Institutions bought selected AI, marketing, and energy exposures while selling selected storage and semiconductor names.

What happens next depends on whether those trades repeat and whether company results justify them. Track the next public seat disclosures, then compare them with earnings, margins, inventories, orders, and price behavior.

That process turns a fleeting newsflash into a testable market view. Until those signals arrive, the most accurate conclusion remains narrow: institutional trading favored Kunlun Tech that day, but the wider technology verdict is still open.