China Computing Stocks Lose Momentum as the A-Share Market Sells Off

RSSHub 36Kr recorded a sharp change in China’s stock market on July 24, with nearly 5,000 companies falling before the midday break. The Shanghai Composite lost 1.2 percent. The Shenzhen Component dropped 1.77 percent, while the ChiNext Index fell 1.78 percent.

Computing infrastructure stocks, previously among the market’s strongest speculative trades, led the retreat. Client-service provider ZST Digital Networks and electronics manufacturer Litong Electronics reached their daily downside limits. MCC Meili Cloud Computing and DHC Software affiliate Beijing Teamsun Technology fell more than 7 percent.

The reversal matters because these companies had benefited from enthusiasm around artificial intelligence infrastructure, data centers, and rented computing capacity. Only two days earlier, several names in the same group had posted strong gains. The latest decline therefore looks less like routine index volatility and more like a rapid reassessment of crowded growth positions.

Banks and telecommunications companies moved in the opposite direction. China Telecom and China CITIC Bank gained more than 1 percent, creating a defensive pocket inside an otherwise broad decline. That rotation offers the clearest explanation of the morning’s market structure, even though it does not establish a single cause.

The immediate conflict is between long-duration technology expectations and businesses valued for current earnings, dividends, or defensive characteristics. Investors did not reject China’s computing expansion outright. They reduced exposure to the most crowded expressions of that expansion while seeking protection elsewhere.

What the RSSHub 36Kr Market Snapshot Actually Shows

The defining feature of the morning was market breadth, not simply the decline in three headline indexes.

The midday market update reported that almost 5,000 stocks were trading lower. That figure shows selling across most of the listed market rather than weakness confined to several large technology companies.

Market breadth measures how many securities participate in a move. A falling index can sometimes hide a healthier market if only a few heavyweight companies decline. That interpretation does not fit this session because losses extended across thousands of companies.

The three major indexes also moved together. The Shanghai Composite declined 1.2 percent, while the Shenzhen Component and ChiNext Index each lost about 1.8 percent. The greater declines in Shenzhen and ChiNext point to heavier pressure on growth-oriented shares.

The retreat was visible from the opening auction. A morning market report showed the Shanghai Composite opening 0.6 percent lower at 3,853.63. The Shenzhen Component opened 1.47 percent lower, and ChiNext started down 1.68 percent.

That opening report also identified weakness in computing hardware, storage chips, optical networking, precious metals, and electronic components. Oil services and gas companies were among the limited areas moving higher.

Several speculative leaders opened under pressure after recent advances. Meili Cloud, ZST Digital Networks, and related computing names had entered the session with considerable short-term momentum. Their early losses signaled that traders were reducing risk before the broader market completed its decline.

By midday, the pressure had intensified. ZST Digital Networks and Litong Electronics reached their permitted daily loss limits. Meili Cloud and Beijing Teamsun Technology declined by more than 7 percent.

These stocks do not represent the entire computing industry. Their businesses, financial profiles, and exposure to artificial intelligence spending differ. Still, traders often group them under themes such as computing rental, cloud infrastructure, data centers, or AI capacity.

That thematic grouping creates a transmission channel. When investors reduce exposure to one high-profile member, they often sell related companies before studying their individual fundamentals. Exchange limits and short settlement cycles can make the process appear abrupt.

The A-share market selloff also extended beyond technology. Nearly 5,000 declining stocks indicate that many investors were cutting overall exposure, not merely exchanging one computing company for another.

The available midday figures do not establish the session’s final outcome. Prices can reverse during the afternoon, and turnover data can change the interpretation. A decline supported by heavy volume carries different implications from one occurring during thin trading.

The snapshot nevertheless establishes three facts. China AI stocks lost leadership, selling spread across most of the market, and defensive financial and telecommunications shares attracted relative demand.

Those facts create the central tension. The market continued to recognize demand for digital infrastructure, yet it punished several listed companies associated with supplying or renting that capacity.

A Crowded Computing Trade Meets a Fast Reversal

The decline tests whether recent computing-stock gains reflected durable earnings expectations or short-term demand for a popular market theme.

On July 22, just two trading days before the reported selloff, the market displayed almost the opposite pattern. A July 22 market review reported that Meili Cloud completed a third consecutive limit-up session.

The same review said ZST Digital Networks and Litong Electronics had each produced two consecutive limit-up sessions. These moves placed the companies among the market’s most visible short-term winners.

That earlier session was not broadly strong. More than 3,800 stocks declined, while the ChiNext Index fell 3.23 percent. Computing rental shares still advanced, showing that speculative capital had concentrated in a narrow group.

July 24 removed that protection. The former leaders fell alongside the wider market, and several experienced the maximum decline permitted during regular trading. The reversal compressed multiple days of positive momentum into one morning.

This pattern matters because market themes often progress through distinct stages. First, investors identify a credible industry trend. Next, capital flows toward direct beneficiaries. Finally, loosely related stocks rise because their names fit the narrative.

Artificial intelligence provides a credible long-term demand story. Model training, inference, data storage, and cloud services require servers, networking equipment, electricity, cooling systems, and suitable facilities.

That demand does not guarantee equal returns for every listed company carrying a computing label. Operators must secure customers, finance equipment, manage energy costs, and keep expensive capacity utilized. Hardware can also lose economic value as newer systems arrive.

Investors therefore face a difficult distinction. They must separate companies with measurable orders and cash generation from companies valued mainly through thematic association.

The July 24 decline does not resolve that distinction. It shows that traders became less willing to pay for uncertainty, especially after rapid short-term gains.

Litong Electronics illustrates the complexity. The company has appeared in market coverage as part of the computing rental theme, but investors still need to examine how much revenue comes from that activity. They must also consider capital requirements and contract durability.

Meili Cloud carries both cloud-related expectations and legacy business considerations. ZST Digital Networks has exposure to financial technology infrastructure and data-center services. Beijing Teamsun Technology participates in enterprise computing through several business lines.

Those differences disappear when a theme becomes crowded. Prices can rise together during enthusiasm and fall together during liquidation. Correlation temporarily replaces company-level analysis.

China AI stocks are especially sensitive to this process because demand expectations stretch across many layers of the supply chain. Chip designers, optical-component makers, server manufacturers, data-center operators, software vendors, and power suppliers can all enter the same narrative.

Yet their economics differ sharply. A component maker with confirmed orders is not equivalent to a rental operator financing new capacity. A telecommunications company with existing networks does not carry the same risk as a smaller provider building facilities ahead of demand.

The correction also followed a period of wider pressure on growth shares. On July 22, ChiNext fell more than 3 percent even as computing rental names advanced. That divergence suggested the theme was already operating against a weakening market backdrop.

Once the thematic bid disappeared, those shares had little protection from the general decline. The result was not simply a bad morning for technology. It was the failure of a narrow leadership group during an A-share market selloff.

There is a historical warning in that pattern. On July 7, Litong Electronics also reached its downside limit during a retreat in computing rental shares. The July 7 session ended with nearly 4,800 stocks lower.

However, July 7 also showed why broad labels require caution. Semiconductor wafers, computing chips, and advanced packaging companies moved higher while rental-related names declined. Even within the AI infrastructure trade, investors distinguished between hardware production and capacity leasing.

The July 24 snapshot provides less evidence of such differentiation. Computing hardware and storage shares opened lower, while rental-related leaders weakened further by midday. That wider retreat raises the possibility of a broader reduction in technology exposure.

It still does not prove that the long-term infrastructure cycle has ended. One morning cannot determine a multiyear capital-spending trend. It does show that momentum alone no longer protected the most crowded stocks.

Banks and Telecoms Become the Other Side of the Trade

The relative strength of banks and telecommunications companies shows investors favoring current cash flows over distant computing expectations.

China Telecom and China CITIC Bank each rose more than 1 percent by the midday break. Their gains were modest in isolation, but they stood out against almost 5,000 declining stocks.

This divergence establishes the session’s primary opponent map. Computing-themed growth shares represented expectations about future demand, while banks and telecoms represented established scale, recurring revenue, and defensive positioning.

The distinction is not absolute. China Telecom is also a major participant in cloud services, data centers, and national computing infrastructure. Its shares can therefore provide exposure to digital demand without relying entirely on a narrowly defined computing rental story.

That dual position helps explain why telecommunications stocks can behave differently from smaller thematic companies. Large operators own networks, customer relationships, and existing infrastructure. Their investment plans sit within broader businesses rather than a single speculative theme.

Banks offer a different form of defense. Their performance depends on credit conditions, margins, asset quality, and the economy. However, investors often treat large banks as relatively stable when high-valuation growth shares face pressure.

The rotation does not mean every investor reached the same conclusion. Some positions may reflect short covering, index rebalancing, dividend preferences, or temporary trading flows. Public midday data cannot identify each buyer’s motivation.

The price pattern still communicates a preference. Investors accepted exposure to mature, regulated businesses while reducing exposure to companies whose valuations depend more heavily on future computing demand.

That preference places smaller infrastructure companies under pressure. They must show that capital spending can produce contracted revenue, acceptable utilization, and sustainable returns. Announcements about capacity are no longer enough when financing costs and technology turnover remain important.

The market’s treatment of China Telecom is particularly revealing. Telecommunications networks are essential to data movement, cloud delivery, and distributed computing. A defensive telecom rally therefore does not contradict the infrastructure thesis.

Instead, it suggests investors were changing how they wanted to own that thesis. They favored an incumbent with diversified operations over smaller companies exposed to short-term thematic flows.

This creates a more useful interpretation than saying investors abandoned AI. The long-term demand argument survived, but risk tolerance changed.

Major infrastructure projects provide evidence that investor interest in the sector remains substantial. Reuters reported that optical-transceiver manufacturer Zhongji Innolight was preparing a large Hong Kong listing after rapid revenue and profit growth.

According to the IPO filing report, Zhongji planned to raise at least $8 billion. Its optical products move data between servers in AI and cloud facilities.

Reuters also reported that the company’s first-quarter revenue reached 19.5 billion yuan, up 192 percent from the prior-year period. Profit increased 274 percent to 6.32 billion yuan.

Those figures show why investors remain interested in China AI stocks. Some suppliers are reporting substantial operating growth tied to real infrastructure demand.

They also sharpen the contrast with less proven computing stories. A company showing large revenue and profit increases offers different evidence from one trading primarily on capacity plans or thematic identification.

Zhongji’s planned listing carries its own risks, including exposure to overseas customers and geopolitical restrictions. It is not a risk-free benchmark. It does provide a useful comparison between measurable operating results and speculative expectations.

The July 24 rotation therefore creates pressure on companies that cannot show equivalent evidence. Investors can pursue the same broad technology trend through profitable suppliers, large telecom operators, or diversified platforms.

Smaller names must compete for capital against all three categories. When markets fall broadly, that competition becomes more demanding.

This is the core reversal. Computing demand remains strategically important, but the market stopped rewarding every company attached to it.

What the Midday Numbers Cannot Prove

The selloff is meaningful, but it cannot by itself establish why investors sold or where prices will finish.

A market report captures observed prices. It does not automatically reveal a causal chain. The RSSHub 36Kr item accurately summarizes the reported midday moves, but its short format leaves several questions unanswered.

First, the snapshot does not provide turnover for the morning. Volume helps analysts judge conviction because intense selling on heavy turnover differs from a decline caused by limited liquidity.

Second, the report does not separate institutional activity from retail trading. China’s mainland market includes significant retail participation, which can increase the speed of thematic rotations.

Third, the report does not identify a single policy announcement, earnings release, or industry disclosure that triggered the computing decline. Broad risk reduction may explain part of the move.

The opening weakness across storage, optical networking, precious metals, and electronic components supports that interpretation. Several unrelated sectors declined together before computing rental shares reached their daily limits.

Fourth, the midday numbers do not show whether listed companies had changed their operating guidance. Without new company disclosures, falling prices reflect a change in market expectations rather than confirmed deterioration in underlying businesses.

That distinction matters. A stock can fall because earlier expectations became too optimistic, even when revenue continues to grow. The correction may target valuation rather than operations.

The available evidence also cannot prove that banks and telecoms will maintain their gains. Defensive rotations can reverse quickly if broader sentiment improves during the afternoon.

Readers should avoid treating a newsflash as a complete trading record. RSSHub is a distribution layer that makes published feeds easier to follow. It does not create an independent set of exchange prices or company disclosures.

Likewise, 36Kr’s newsflash format prioritizes speed and compression. It tells readers what moved and by how much. It does not replace filings, exchange announcements, or a full market-data terminal.

The phrase RSSHub 36Kr can therefore describe how readers discovered the update, but it should not be confused with a market index. The underlying evidence comes from reported A-share prices at the midday break.

There is another uncertainty around sector classification. “Computing stocks” can refer to data-center operators, computing rental providers, server makers, optical-component suppliers, chip companies, or cloud software businesses.

Combining them can obscure material differences. A provider with long-term customer contracts faces different risks from a company buying hardware before securing demand.

Power availability also matters. Data centers require reliable electricity, cooling, land, and network access. Companies with capacity on paper may not convert it into economical, usable computing services.

Hardware depreciation adds another pressure. Accelerators and servers can become less competitive as newer systems improve performance or energy efficiency. Operators must earn sufficient returns before equipment loses value.

Customer concentration creates additional uncertainty. A computing provider dependent on a few buyers may report fast growth but still carry significant renewal risk.

Financing structure matters as well. Debt-funded expansion can magnify returns when utilization remains high. It can also magnify losses when demand arrives later than expected.

None of these risks proves that the companies falling on July 24 face the same problems. Their latest filings would need to be examined individually.

The sector’s recent price behavior does, however, show that investors are questioning how those risks should be valued. Repeated limit moves in both directions indicate uncertainty rather than a settled fundamental judgment.

An A-share market selloff can amplify that uncertainty. Traders facing losses elsewhere may sell their most liquid winners, regardless of their long-term outlook. That process can turn crowded leadership into a source of cash.

The broad decline also makes it difficult to isolate a computing-specific signal. When almost 5,000 stocks fall, company fundamentals explain only part of the day’s movement.

Investors should therefore resist two extreme conclusions. The selloff does not prove that China’s AI infrastructure expansion has failed. It also does not guarantee that every falling computing stock will recover.

The more defensible conclusion is narrower. Momentum reversed, market breadth deteriorated, and investors demanded a stronger margin of safety from high-expectation technology shares.

That interpretation matches the observable evidence without inventing a cause. It also creates clear tests for the weeks ahead.

Why China AI Stocks Still Face a Valuation Test

The market is shifting from asking who can announce computing capacity to asking who can earn acceptable returns from it.

China’s capital-market policy continues to support technology development and advanced manufacturing. A 2025 capital market policy directed financial resources toward technological innovation and digital finance.

That policy background supports long-term investment, but it does not remove company-level risk. Public policy can increase available capital without ensuring that every recipient builds a profitable business.

Computing infrastructure is particularly vulnerable to this distinction. National demand can grow while individual operators struggle with utilization, financing costs, procurement, or price competition.

The market’s July 24 behavior suggests investors are applying a stricter filter. Companies with visible orders, defensible technology, and growing profits retain a stronger case than companies dependent on narrative momentum.

Zhongji Innolight provides one benchmark because its recent growth connects directly to optical components used in data centers. China Telecom provides another because it combines infrastructure exposure with a diversified operating base.

Smaller rental-oriented companies face a harder comparison. They must demonstrate that installed capacity is being used by paying customers and that contract revenue covers equipment, energy, and financing expenses.

That does not make large companies automatically superior. Large telecom operators can face slower growth, heavy capital spending, and regulatory constraints. Component suppliers can face customer concentration and geopolitical risks.

The relevant difference is evidence. Investors can compare reported revenue, profit, customers, and investment commitments rather than relying only on thematic labels.

This evidence-based approach also helps explain why one technology subgroup can fall while another rises. On July 7, semiconductor companies advanced even as computing rental shares retreated. The market separated hardware manufacturing from leasing economics.

On July 24, the selloff appeared broader, but that earlier distinction remains useful. “AI infrastructure” is not a single business model.

For developers and enterprise buyers, the stock correction does not directly change access to computing services. It may still influence which providers can finance expansion and how aggressively they compete for customers.

A well-capitalized provider can purchase newer hardware, secure energy, and expand network connections. A company under market pressure may slow investment or seek partnerships.

That connection makes the selloff relevant beyond trading desks. Public-market confidence can affect the pace and ownership structure of infrastructure development.

Knowledge workers and AI product teams should care for another reason. The cost and availability of computing capacity shape model access, service reliability, and the economics of AI applications.

Yet users should not infer immediate service disruptions from falling share prices. No evidence in the midday report indicates that data centers stopped operating or that customers lost access.

The market is evaluating expected future returns, not reporting a present infrastructure failure. This difference should remain central to any interpretation of China AI stocks.

The valuation test will become more demanding if defensive sectors continue outperforming. Investors then have less incentive to tolerate uncertain cash flows from smaller technology companies.

It will become less demanding if computing companies publish stronger contracts, higher utilization, or better profits. Verified operating performance can restore confidence even after a momentum-driven correction.

The next phase therefore depends less on slogans about national computing power. It depends on financial statements and operational disclosures.

That is why the July 24 move is more than a routine red screen. It marks a contest over which evidence investors now require before funding the next stage of infrastructure growth.

Three Signals to Watch After the A-Share Market Selloff

Afternoon breadth, company disclosures, and the durability of sector rotation will determine whether the morning was a reset or a deeper repricing.

The first signal is the final market breadth on July 24. If close to 5,000 stocks remain lower at the closing bell, the session will represent sustained market-wide risk reduction.

A meaningful afternoon recovery would weaken that interpretation. It would suggest the midday figures captured a temporary wave of selling rather than a settled change in risk appetite.

Turnover must accompany that closing assessment. Heavy trading during continued declines would indicate stronger conviction. Lower turnover would support a more cautious reading.

The second signal is operating evidence from computing-related companies over the next reporting cycle. Investors should examine revenue from computing services, customer concentration, contract duration, utilization, capital spending, and cash flow.

These figures will determine whether the decline primarily removed speculative excess or anticipated weaker business conditions. Companies with growing revenue but deteriorating cash flow will require especially close attention.

Investors should also watch whether management teams distinguish signed customer demand from planned capacity. Announced investment does not produce revenue until infrastructure is installed, connected, and used.

The third signal is whether banks and telecommunications shares keep outperforming. One defensive morning can reflect short-term positioning. Several weeks of relative strength would show a more persistent preference for established cash flows.

China Telecom deserves particular attention because it sits on both sides of the market’s debate. It offers defensive characteristics while also investing in cloud and digital infrastructure.

Continued telecom strength alongside weakness in smaller rental providers would reinforce the article’s central judgment. Investors would still want computing exposure, but through diversified operators.

A reversal would weaken that judgment. If smaller providers recover leadership while banks lag, traders may have treated July 24 as a temporary liquidation event.

The same test applies to listed hardware suppliers. Strong performance from profitable optical, server, or semiconductor companies would show that investors are differentiating business models rather than abandoning infrastructure.

RSSHub 36Kr provided an efficient view of the midday damage, but the next stage requires more than a headline. Readers should compare the closing tape with verified company disclosures and sector-relative performance.

The useful question is not whether computing demand still exists. Evidence from data-center investment, telecommunications networks, and component growth shows that it does.

The question is which listed companies can convert that demand into durable earnings before equipment costs, competition, and financing pressure absorb the returns.

Watch those three signals over the next one to three months. They will show whether July 24 was a brief momentum unwind or the start of stricter valuation discipline across China’s computing trade.