Demingli Leads Institutional Buying as GigaDevice Faces a ¥808 Million Exit

Demingli drew ¥513 million in reported institutional net buying on August 3, while GigaDevice faced ¥808 million in net selling. An RSSHub 36Kr feed surfaced the figures, which originated from a First Financial summary of after-hours trading disclosures.

The split matters more than the headline totals. Both companies participate in China’s semiconductor economy, yet large trading desks moved in opposite directions. Institutions favored selected memory and materials names while reducing exposure to another prominent chip supplier.

The figures came from China’s Dragon and Tiger List, an after-hours disclosure of leading trading seats in unusually active securities. It offers a narrow view of exceptional trading, not a complete record of institutional portfolios.

That limitation creates the central tension. The list shows where disclosed institutional seats traded aggressively, but it does not explain their strategies. Investors must separate a useful flow signal from a durable judgment about company value.

The August 3 List Revealed a Sharp Institutional Split

The strongest signal was not broad confidence in technology stocks, but concentrated buying and even more concentrated selling.

Institutions appeared in 25 stocks on the August 3 list, according to the reported after-hours summary. Ten recorded institutional net buying, while 15 recorded institutional net selling.

Demingli ranked first among the net purchases at ¥513 million. Jacques Technology followed at ¥273 million, and Leo Group ranked third at ¥146 million.

The leading sales were larger. GigaDevice recorded ¥808 million in institutional net selling, while Honghe Technology registered ¥564 million. Shunna Electric followed with a reported ¥73.09 million outflow.

Those totals describe net activity attributed to institutional seats. Net buying subtracts disclosed sales from disclosed purchases, so it does not equal the gross value of all orders.

The difference is important because one security can contain institutional seats on both sides. A positive result means the disclosed buying exceeded disclosed selling, not that every institution agreed.

Likewise, the 25-stock count should not be treated as a market-wide institutional survey. Securities enter the list after meeting specific disclosure conditions tied to unusual trading activity.

The list normally identifies the leading buying and selling seats for each qualifying security. It therefore emphasizes the most active visible participants rather than every investor involved.

That structure makes the data valuable as an event signal. It shows where unusual price action intersected with identifiable institutional trading after the close.

It also makes the data incomplete. Positions accumulated before August 3, trades outside the disclosed seats, and hedges in related securities remain outside the snapshot.

The imbalance between GigaDevice and Demingli still deserves attention. GigaDevice’s reported institutional outflow exceeded Demingli’s inflow by ¥295 million, despite both companies’ exposure to semiconductor demand.

Honghe Technology’s ¥564 million outflow was also larger than the purchases recorded for every company except Demingli. The day therefore reflected selective rotation, not simple enthusiasm for growth shares.

That distinction changes how readers should interpret the event. Institutions were not uniformly buying chips, storage, or technology. They were separating companies within those categories.

Why Demingli Attracted the Largest Reported Inflow

Demingli’s position at the top of the buying list fits a broader market focus on memory demand, but one session cannot confirm the underlying thesis.

Demingli develops storage products, including solid-state drives, embedded storage, and memory modules. This places the company near demand from consumer devices, enterprise systems, and data-intensive computing.

Its annual filing described expanding enterprise storage activity and continued customer validation. The filing also linked stronger storage demand with artificial intelligence infrastructure.

The company reported 2025 embedded-storage revenue of about ¥3.66 billion, up 334.43 percent. Memory-module revenue reached roughly ¥1.05 billion, an increase of 263.65 percent.

Those historical results provide context for investor interest. They do not establish why any disclosed seat bought shares on August 3.

Memory businesses can benefit when demand rises faster than available supply. Higher component prices can lift revenue, inventory values, or margins for some participants.

The same cycle can introduce risk. Companies may pay more for flash and memory components, carry expensive inventory, or encounter weaker demand after ordering aggressively.

Demingli’s filing said fourth-quarter spot indexes for both DRAM and NAND rose by more than 150 percent. It attributed those figures to data from CFM Flash Market.

That price environment gives the market a plausible reason to revisit storage suppliers. It also raises the stakes around procurement, inventory discipline, and customer acceptance.

Enterprise storage offers another possible attraction. Artificial intelligence servers and data centers require substantial capacity for models, training data, retrieval systems, and operational records.

However, rising data volumes do not benefit every storage company equally. Product mix, controller design, qualification cycles, component access, and customer concentration all affect the outcome.

The August 3 inflow therefore looks consistent with a memory-cycle thesis. It does not prove that institutions expect a specific earnings result or product milestone.

The distinction matters because the Dragon and Tiger List does not identify a single unified decision-maker. Several institutions can trade through separate designated seats for unrelated reasons.

One fund may be building a longer position. Another may be covering a short exposure, rebalancing an index strategy, or reacting to a rapid price move.

Demingli had also appeared in earlier periods of intense market activity. A May 13 market summary reported that it led institutional net buying on that day as well.

Repeated appearances can show sustained trading interest. They can also indicate unusually high volatility, which is one reason the exchange publishes the underlying seat information.

Investors should therefore connect the flow with operating evidence. The useful questions concern shipments, margins, inventory turnover, and enterprise customer conversion.

Until those indicators confirm the story, ¥513 million remains a significant trading event rather than a complete investment case.

GigaDevice’s ¥808 Million Outflow Is the Real Reversal

The day’s central reversal was institutional enthusiasm for one memory-linked company alongside heavier selling in a larger semiconductor name.

GigaDevice supplies memory products, microcontrollers, sensors, and related semiconductor solutions. That broader portfolio gives it exposure to consumer electronics, industrial systems, automotive applications, and connected devices.

Its position in several chip categories makes the outflow harder to reduce to one simple narrative. The selling might reflect valuation, portfolio construction, earnings expectations, or short-term risk control.

It might also reflect profits being taken after an earlier advance. A disclosed sale does not automatically mean an institution rejected the company’s long-term technology position.

The size nevertheless stands out. At ¥808 million, GigaDevice’s reported net institutional selling was approximately 57.5 percent larger than Demingli’s reported net buying.

This creates a stronger signal than Demingli’s purchase total alone. Capital did not merely enter a favored storage stock. A larger amount left another widely followed semiconductor company through disclosed institutional seats.

The difference suggests investors were distinguishing between exposures within the chip sector. They were not treating semiconductor demand as a single trade.

Memory-module assemblers, flash designers, microcontroller suppliers, and materials companies occupy different positions in the supply chain. Each faces distinct pricing and inventory dynamics.

A component-price increase can improve conditions for one company while pressuring another. Inventory acquired at lower prices can become valuable, while expensive replenishment can reduce future margins.

Product cycles also move at different speeds. Enterprise storage demand can strengthen while consumer electronics remain uneven. Industrial and automotive orders can follow separate replacement patterns.

These differences help explain why Demingli and GigaDevice can move in opposite institutional directions. The list does not provide enough evidence to identify which factor dominated.

Honghe Technology’s ¥564 million outflow adds another semiconductor-related counterpoint. The company produces electronic-grade glass fiber materials used in circuit-board supply chains.

Jacques Technology, meanwhile, drew ¥273 million in net institutional buying. Its businesses include materials serving semiconductor manufacturing and other advanced industrial applications.

The paired outcomes reinforce the selection pattern. Institutions bought some hardware-chain exposures while selling others, even within adjacent areas of electronics production.

That is the opposite of an indiscriminate sector rally. It resembles a rotation toward specific earnings sensitivities, supply positions, or perceived valuation opportunities.

Still, the transaction list cannot reveal whether the same institutions sold GigaDevice and bought Demingli. Treating the two totals as a direct swap would overstate the evidence.

The list also cannot show portfolio weights after the trades. An institution selling part of a large holding might remain heavily invested in GigaDevice.

This is why the outflow should be framed as pressure, not abandonment. It establishes an exceptional trading event that requires confirmation from later sessions and company disclosures.

What the Dragon and Tiger List Does Not Show

Institutional-seat data offers transparency around unusual activity, but it cannot identify a durable consensus or predict the next price move.

China’s exchanges publish trading details when securities satisfy defined abnormal-trading conditions. The resulting list commonly shows the five leading buying and selling seats.

“Dedicated institutional” is a seat classification. It can represent activity associated with funds, securities firms, insurers, social-security capital, and other professional investors.

That label does not disclose the final beneficial owner in a simple public name. Readers cannot assume one institution produced the entire net figure.

The list also covers only qualifying securities. A stock with large institutional activity may remain absent if it does not meet the relevant disclosure threshold.

This selection effect makes cross-market conclusions unreliable. Twenty-five disclosed stocks do not represent every institutional decision made in mainland equities that day.

Time horizon presents another problem. The same transaction can mean different things depending on whether the trader expects to hold for hours, weeks, or years.

A quantitative strategy may respond to price and liquidity signals. An active fund may rebalance around risk limits. A fundamental manager may react to new operating evidence.

Each action can appear as institutional buying or selling. The public data does not provide the investment memorandum behind it.

Net figures also hide gross activity. An institutional seat that bought ¥500 million and sold ¥400 million produces only ¥100 million in net buying.

Another stock might show the same net amount from ¥100 million in buying and no disclosed selling. Those two trading patterns carry different information.

Price context matters as well. Net buying into a sharp rise can indicate conviction, momentum exposure, or late participation. Buying into weakness can reflect accumulation or an attempted stabilization.

Without the full price path, turnover, disclosure trigger, and seat details, the headline amount remains only one layer of evidence.

There is also no guarantee that institutional flows lead future performance. Professional investors can disagree, mistime cycles, or change positions quickly when new information arrives.

The list’s value lies in disciplined use. It identifies securities deserving closer examination and shows where professional trading intersected with unusual market behavior.

The list becomes misleading when readers treat it as a recommendation service. A large purchase is not equivalent to verified earnings growth, and a large sale is not proof of deterioration.

For Demingli, the next step is to compare the flow with storage shipments, revenue quality, and inventory. For GigaDevice, readers need operating trends and management disclosures.

This evidence-first approach also matters when following aggregated alerts. An RSSHub 36Kr item can deliver the event quickly, but the alert should begin verification.

Investors should return to exchange disclosures, company filings, and subsequent financial results. Those sources offer the context required to test a trading narrative.

The Semiconductor Supply Chain Is Not One Trade

August 3 exposed competing sensitivities across storage products, chip design, semiconductor materials, and electronics manufacturing.

Demingli’s reported inflow placed memory systems at the center of the buying story. Its products depend on access to flash and DRAM components, product qualification, and customer demand.

GigaDevice operates through a different mix. Its portfolio spans flash memory, microcontrollers, sensors, and analog-related products, creating broader but more complex demand exposure.

Jacques Technology adds a materials perspective. Its businesses connect to manufacturing inputs rather than the same product layer occupied by storage-device suppliers.

Honghe Technology represents another upstream segment through electronic-grade glass fiber materials. Such materials support printed circuit boards used across electronics systems.

These companies can all appear in a “technology” or “semiconductor” basket. Their earnings drivers remain materially different.

Storage suppliers can benefit from unit growth, favorable procurement, and improving product mix. They can suffer when component costs rise faster than selling prices.

Chip designers depend on product competitiveness, customer inventory, wafer access, and end-market demand. Their cycles can diverge across consumer, industrial, and automotive applications.

Materials companies depend on factory utilization, customer qualification, supply contracts, and manufacturing expansion. Their response to a chip upcycle can arrive at a different time.

The August 3 flows therefore look more like a supply-chain selection exercise than a simple bet on artificial intelligence. AI demand provides context, not a complete explanation.

Artificial intelligence infrastructure increases requirements for computation, memory, storage, networking, and power. However, revenue capture depends on each supplier’s position and execution.

Demingli’s reported enterprise-storage expansion gives investors one operating narrative to examine. The company still needs to translate demand into sustainable revenue and cash generation.

GigaDevice’s broad catalog gives it several possible growth routes. The same breadth can expose it to mixed inventory conditions across customer categories.

Jacques Technology’s institutional inflow suggests buyers also considered manufacturing materials attractive. That move does not establish a single shared reason across all purchasing seats.

Leo Group complicates the picture further. Its reported ¥146 million institutional inflow came from a company with digital-marketing and industrial operations, not a pure semiconductor profile.

The broader list therefore contained multiple market stories. Grouping all ten net purchases under one technology thesis would erase meaningful differences.

The ratio of buyers to sellers also warns against a broad bullish interpretation. Fifteen stocks recorded net selling, compared with ten showing net buying.

The largest sale exceeded the largest purchase. The second-largest sale also exceeded every reported purchase except Demingli’s.

Those comparisons indicate defensive repositioning alongside targeted accumulation. Institutions appeared willing to concentrate capital, but they also reduced exposure forcefully elsewhere.

For technology investors, the useful unit of analysis is the individual earnings mechanism. Sector labels are too broad to explain why money entered one company and left another.

That is especially true during volatile component cycles. Pricing, inventories, and customer demand can transfer gains and pressure between adjacent companies.

A supply shortage can improve selling prices but restrict available units. An inventory build can protect shipments but increase balance-sheet risk if prices later fall.

The winners are determined by contracts, timing, product quality, and financial discipline. A one-day institutional-flow list cannot settle those questions.

Three Signals Will Test the Institutional Rotation

The August 3 pattern becomes meaningful only if later trading and operating data confirm that institutions favored specific fundamentals rather than temporary volatility.

The first signal is persistence in disclosed flows. Investors should watch whether Demingli continues attracting institutional buying across later abnormal-trading disclosures.

Repeated buying would strengthen the view that professional investors are building exposure to its storage thesis. A quick reversal would weaken that interpretation.

GigaDevice deserves the same test in the opposite direction. Further institutional selling would indicate sustained pressure, while renewed buying would suggest a temporary rebalance.

The second signal is company-level operating evidence. Demingli’s future reports should show whether enterprise storage and other products continue converting demand into shipments.

Revenue growth alone will not answer that question. Gross margin, operating cash flow, inventory, receivables, and customer concentration provide essential context.

A rising inventory balance can support future sales when components are scarce. It can also become a liability when market prices decline.

Cash flow can help distinguish profitable expansion from growth financed by working-capital demands. Receivables can show whether sales are turning into collected cash.

For GigaDevice, investors should examine demand across memory, microcontrollers, and other product lines. Segment trends can reveal whether the outflow anticipated operational pressure.

The third signal is the direction of the memory cycle. Spot prices, contract pricing, production decisions, and data-center demand will shape results across the supply chain.

Demingli’s storage outlook emphasized AI-linked demand and higher component prices. Later disclosures must show how those conditions affected margins and inventory.

If memory demand remains firm while enterprise products gain customers, the August 3 buying will look more consistent with improving fundamentals.

If component prices outrun customer demand, the same inventory exposure can become a risk. That outcome would weaken the bullish interpretation of the flow.

These three tests should be applied in sequence. First establish whether the trading pattern persisted, then examine company results, and finally place those results within the component cycle.

The sequence prevents a familiar analytical error. Investors often start with a dramatic flow figure and search afterward for facts that support it.

A better method treats the figure as a hypothesis generator. The ¥513 million inflow asks whether Demingli’s storage position is improving.

The ¥808 million outflow asks whether GigaDevice faces company-specific pressure, valuation concerns, or short-term portfolio adjustment.

Neither question has a final answer in the August 3 list. The disclosure identifies the conflict but leaves its duration and cause uncertain.

For readers tracking the event through an RSSHub 36Kr feed, the next useful action is not chasing the largest number. It is building an evidence timeline.

Record later institutional-seat disclosures, the companies’ next financial updates, and changes in memory pricing. Compare those signals before assigning a lasting narrative.

Will Demingli’s operating results validate the disclosed buying, or will GigaDevice’s broader portfolio recover institutional support first? The next reports should decide which signal mattered.