China's STAR Market Midyear Results Put Hard-Tech Profits Ahead of Expansion

China's STAR Market posted a striking midyear shift: profits among early reporters rose 136%, far ahead of their 29% revenue growth. The figures appeared in a 36Kr newsflash distributed through an RSSHub 36Kr feed, but the underlying event is bigger than its delivery channel.

By August 25, 251 STAR Market companies had released first-half reports. Together, they recorded RMB 447.7 billion in revenue and RMB 54 billion in net profit, according to the midyear earnings data carried by 36Kr.

That gap between revenue and profit growth is the important part. It suggests that at least some Chinese hard-tech businesses gained operating leverage after years of research spending, capacity building, and supply-chain localization.

Yet this is not a simple victory over the industrial cycle. Semiconductor demand, AI infrastructure orders, loss reversals, and a small number of large contributors can all lift an aggregate quickly. The midyear reports therefore test two competing explanations.

One says hard-tech companies are finally converting research into durable earnings. The other says favorable demand, capacity utilization, and comparison effects temporarily amplified the numbers.

The final disclosure set strengthens the headline while also sharpening that tension. After all Shanghai-listed companies completed their reports, the Shanghai Stock Exchange said the entire STAR Market generated RMB 1.01 trillion in revenue and RMB 144.887 billion in profit.

Those totals represented growth of 38.6% and 437.6%, respectively. They show that the pattern continued beyond the first 251 reporters. They do not, by themselves, prove that every company or technical field shared equally in the gains.

The RSSHub 36Kr Headline Captured Only an Early Snapshot

The initial figures were real, but they described less than half of the market and arrived before the reporting season closed.

The August 25 snapshot covered 251 companies, or slightly more than 40% of the STAR Market. Among them, 195 were profitable, 111 increased profit, and 28 moved from losses to earnings, according to a separate reporting-season update.

These details matter because aggregate profit growth can come from several different mechanisms. Existing profitable companies can expand margins. Previously unprofitable businesses can cross break-even. Large companies can contribute a disproportionate share of the increase.

A company moving from a loss to even a modest profit produces an unusually large change in the aggregate. That improvement is economically meaningful, but its percentage growth does not behave like growth from a stable earnings base.

The RSSHub 36Kr item also arrived during a fast-moving reporting window. Six days earlier, only 88 companies had reported. That group had produced RMB 204.3 billion in revenue and RMB 18.9 billion in profit.

Revenue for those 88 companies increased 32%, while profit rose 154%. The similarity to the August 25 pattern suggested the earnings surge was not created by one late batch of filings.

However, the expanding sample changed the scale and composition of the totals. It added businesses from different industries, with different cost structures, reporting calendars, and exposure to global demand.

By August 30, the completed market review covered the whole STAR Market. Revenue reached RMB 1.01 trillion, while net profit reached RMB 144.887 billion.

The full-market growth rates were even higher than the early snapshot. Revenue rose 38.6%, and profit increased 437.6%. Net profit also exceeded the market's total for the previous full year, according to the exchange.

That result makes the broad recovery harder to dismiss. Still, it introduces an important analytical problem. A larger final percentage does not automatically mean the recovery became more evenly distributed.

The completed report included companies with enormous earnings contributions. It also reflected reversals from weak or negative prior-period comparisons. Investors need both aggregate totals and company-level distributions to understand the quality of that growth.

The exchange reported a median gross margin of 36.8% for STAR Market companies. Research and development spending reached RMB 104.4 billion, up 14.6%, while median research intensity stood at 12.6%.

Those figures offer better support for the innovation thesis than profit growth alone. They indicate that companies did not collectively produce the earnings surge by abandoning research investment.

Research spending nevertheless remains an input, not proof of future returns. Its commercial value depends on whether products win customers, achieve acceptable yields, and generate recurring orders without permanent subsidies or uneconomic capacity expansion.

The final results therefore change the basic question. The issue is no longer whether STAR Market profits improved. The filings establish that they did.

The question is how much of that improvement came from repeatable technical and commercial gains. Answering it requires looking below the market total and into the industries driving the change.

Semiconductor Economics Moved From Drag to Earnings Engine

Semiconductors supplied the strongest evidence that higher utilization and localized production can convert technical investment into operating leverage.

The integrated-circuit chain reported RMB 125.267 billion in first-half profit, according to the Shanghai Stock Exchange. That represented a 660.3% increase from the comparable period.

Its median gross margin reached 33.5%. The exchange attributed the performance to gains across design, manufacturing, equipment, materials, and advanced packaging, rather than one isolated link.

This breadth matters because semiconductor supply chains operate as connected systems. A domestic chip designer still depends on fabrication, testing, production equipment, materials, and software support.

A bottleneck in any one layer can constrain output elsewhere. Conversely, rising orders at foundries can pull equipment, materials, and packaging suppliers into the same expansion cycle.

SMIC and Hua Hong recorded their highest quarterly revenue levels, according to the exchange. That is evidence of stronger foundry demand and better capacity use, although the review did not separate every contributor to the records.

Higher utilization can produce a sharp change in foundry earnings. Depreciation, plant operations, engineering staff, and maintenance create substantial fixed costs. When more wafers move through the same production base, those costs are spread across more output.

The effect can make profit rise far faster than sales. It also works in reverse when demand weakens, which is why one strong half cannot settle the durability question.

Equipment makers offered a second mechanism. Advanced equipment can turn a long research cycle into recurring deliveries, service revenue, and a stronger position in customer production lines.

The exchange said AMEC's 12-inch etching systems were being used for several steps involving devices at three nanometers and below. It also said Piotech equipment had supported roughly 600 million cumulative wafer passes on customer lines.

Those are more informative signals than broad claims about technological self-reliance. Customer-line use indicates that equipment has moved beyond laboratory development and into production environments.

Production qualification creates switching costs because chip manufacturers must validate repeatability, contamination control, yield, and maintenance performance. A qualified supplier can therefore gain a more defensible revenue base.

Materials companies benefited from the same expansion. The exchange said PERIC Special Gases nearly tripled revenue from tungsten hexafluoride, a gas used in semiconductor manufacturing.

National Silicon Industry Group affiliate ESWIN Materials exceeded monthly capacity of one million 12-inch silicon wafers, according to the review. Shenghejing Microelectronics also moved a micro-bump-based 3D integration platform into mass production.

Together, these examples support a mechanism-based explanation. Companies invested in capabilities that customers subsequently used at commercial scale, while higher industry demand improved capacity utilization.

However, they do not remove the cycle from semiconductor economics. Memory pricing, inventory replenishment, AI server construction, consumer electronics demand, and capital spending still influence orders.

The original hard-tech earnings account described the semiconductor industry as entering a strong cycle. That framing is compatible with localization, not an alternative to it.

Domestic substitution can increase a supplier's share while a global upturn expands the entire market. Both forces can operate simultaneously, making it difficult to assign profit growth to one cause.

A durable shift would appear in stable margins and repeat orders after the easiest utilization gains pass. It would also show customers using domestic products without relying mainly on emergency sourcing or policy-driven procurement.

This distinction separates technical qualification from commercial resilience. Qualification opens the door. Sustained yield, service, delivery, and product evolution keep the supplier inside.

AI Infrastructure Is Pressuring Buyers to Validate Domestic Systems

China's AI infrastructure demand is turning domestic computing hardware from a policy objective into a large-scale deployment test.

The Shanghai Stock Exchange grouped four representative STAR Market computing companies, including Hygon Information Technology and Cambricon. Together, they generated RMB 18.155 billion in first-half revenue, up 82.2%.

The exchange said their core products had been adapted for mainstream domestic large language models. It also pointed to work on scale-up nodes and clusters containing thousands of accelerators.

An accelerator cluster combines processors, memory, networking, software, and cooling into one computing system. Performance depends on the interaction among those components, not merely the advertised capability of a chip.

That makes AI infrastructure a demanding test for China's hard-tech earnings story. Customers need systems that can train or serve models reliably, fit existing software workflows, and deliver acceptable economics under sustained use.

A chip vendor can record rapid revenue growth during an infrastructure buildout. Long-term value requires the customer to keep using the platform after initial procurement and testing.

Software compatibility is especially important. Developers need compilers, libraries, model frameworks, debugging tools, and stable documentation. Missing software can leave expensive hardware underused.

The exchange's wording that products had completed adaptations for domestic models is therefore relevant. It suggests movement toward usable systems, but it does not establish parity across every workload or deployment size.

Large clusters introduce further constraints. Interconnect performance, power distribution, thermal management, memory bandwidth, and failure recovery become harder as systems grow.

A deployment that performs well in a small test does not automatically retain the same efficiency across thousands of devices. Customers must validate throughput, utilization, stability, and operating cost in production.

This places pressure on more than accelerator vendors. Server makers, networking suppliers, printed circuit board companies, optical component producers, data-center operators, and software teams all affect the final result.

The exchange said Shengyi Electronics and Founder Technology increased net profit by 109% and 232%, respectively, amid demand for supporting components and higher-end capacity.

These gains show how AI spending can spread through the hardware chain. They also highlight concentration risk, because several suppliers can depend on the same small group of large infrastructure buyers.

An order surge from major cloud, telecom, or public-sector customers can lift a supplier rapidly. Delayed projects or weaker capital budgets can reverse that lift just as quickly.

The strongest interpretation is therefore not that domestic AI hardware has conclusively displaced foreign platforms. The filings show that domestic systems reached broader commercial deployment and generated substantial revenue growth.

The unresolved issue is workload quality. Investors and enterprise buyers need evidence about sustained utilization, software migration costs, energy efficiency, failure rates, and repeat purchasing.

This is where the contest between innovation-led earnings and cycle-led earnings becomes concrete. Repeat orders following production use would support the first explanation.

Orders driven mainly by initial capacity construction would leave the second explanation alive. The next reporting periods should reveal whether revenue follows actual computing consumption or primarily reflects hardware installation.

The same distinction applies to end-user software. The exchange said Kingsoft Office's enterprise AI subscription revenue had grown more than 60% for six consecutive quarters.

That streak indicates recurring adoption beyond a single hardware shipment. Yet it covers one business and cannot stand in for the entire domestic AI stack.

For knowledge workers, the implications depend on what these investments produce at the application layer. More local computing capacity matters when it improves model availability, response time, data controls, or workflow economics.

For enterprise buyers, the practical task is to track deployment evidence rather than slogans. Useful evidence includes renewal rates, system utilization, software support, application performance, and total operating requirements.

The earnings reports show capital and customer demand moving toward domestic AI systems. The next phase must show that those systems remain productive after installation.

What the Profit Surge Does Not Prove

The market totals show a powerful recovery, but concentration, comparison effects, and cyclical exposure limit what can be inferred from one half-year.

The largest caution comes from the semiconductor total itself. Integrated-circuit companies generated RMB 125.267 billion of profit, while the entire STAR Market generated RMB 144.887 billion.

Those reported figures indicate that integrated circuits accounted for most aggregate earnings. They also show why the market-wide growth rate should not be read as a uniform result across every hard-tech category.

One company illustrates the concentration issue. The exchange said memory producer ChangXin Technology earned RMB 77.6 billion and moved from a prior-period loss to profit.

That contribution represented more than half of the integrated-circuit profit stated by the exchange. A large turnaround can transform the total even when many smaller companies experience more modest changes.

This does not make the earnings unreal. It changes their interpretation. Aggregate growth describes the sum, while investors often need the median company's operating trajectory.

The August 25 sample showed 195 profitable companies out of 251. It also showed 111 with profit growth and 28 turnarounds.

Those counts confirm that improvement extended beyond one business. Yet they also imply that not every reporting company increased earnings or produced a profit.

The full-market exchange review gave a median gross margin and research intensity, but it did not provide a complete English-language distribution of revenue growth, profit growth, or cash conversion by company.

Cash flow offers another pressure test. Across Shanghai's nonfinancial real-economy companies, operating cash flow reached RMB 1.52 trillion, up 35.1%, according to the exchange.

That total covered a broader group than the STAR Market, so it cannot directly validate every hard-tech company's earnings. Company filings remain necessary for checking receivables, inventory, contract liabilities, and capital spending.

Inventory deserves particular attention in semiconductors. Rising inventory can prepare a company for demand, but it can also indicate slower sales or products at risk of price erosion.

Receivables deserve similar scrutiny. Revenue growth supported by delayed customer payments can produce weaker cash economics than the income statement suggests.

Government support can also affect results through grants, tax treatment, procurement, or financing conditions. Such support can be strategically rational while still complicating comparisons with businesses operating under different conditions.

Another uncertainty concerns capital intensity. The exchange said Shanghai's emerging industries maintained high investment, while cash paid for long-term asset construction across relevant companies reached RMB 406.6 billion, up 4.4%.

New capacity can support future growth. It can also pressure returns if demand disappoints or technology changes before plants reach efficient utilization.

Hard-tech businesses face unusually long feedback loops. A fabrication plant, manufacturing tool, medical platform, or advanced material can require years of investment before commercial acceptance.

That lag explains why profit growth can arrive suddenly after a long development period. It also explains why one period of strong earnings cannot establish a stable return on invested capital.

The international environment adds another source of uncertainty. Export controls, procurement rules, equipment access, and cross-border customer decisions can alter both demand and production capability.

Localization can protect suppliers from some external restrictions. At the same time, restrictions can raise development costs or limit access to specialized tools.

The most defensible conclusion is narrower than the bullish headline. China's STAR Market produced a broad and unusually large earnings improvement during the first half of 2026.

Semiconductor manufacturing, equipment, materials, advanced packaging, and AI infrastructure all supplied concrete commercial examples. Research spending also continued to rise.

However, the results do not prove that industrial cycles have stopped mattering. They do not show that all companies share the same economics, or that current margins will survive weaker demand.

The 2026 figures mark a shift from financing technical ambition toward measuring commercial output. That is progress, but it also raises the standard of evidence expected from the next reports.

Three Signals Will Test Whether Hard-Tech Growth Lasts

Repeat demand, cash conversion, and margin stability will determine whether this earnings wave reflects a structural transition or an unusually favorable period.

The first signal is order quality across semiconductors and AI infrastructure. Investors should watch whether foundries, equipment makers, materials suppliers, and accelerator vendors report repeat orders after current installations enter production.

A repeat order carries more information than an initial qualification. It suggests that the product met customer requirements for yield, reliability, service, or computing performance.

For AI systems, the strongest evidence would connect hardware delivery with rising production workloads. That includes continued software adaptation, measurable utilization, and customer expansion beyond pilot clusters.

If repeat orders remain strong across multiple customer groups, the innovation-led explanation gains credibility. If orders slow after the current capacity wave, cyclical demand becomes the stronger account.

The second signal is cash conversion. The next filings should show whether reported earnings produce operating cash flow without an outsized buildup in receivables or inventory.

Strong cash conversion would indicate that customers are paying and products are moving through the channel. Weak conversion would raise questions about sales quality, bargaining power, or capacity built ahead of demand.

This measure needs company-level review. Market-wide cash flow can hide sharp differences between profitable leaders, early-stage developers, and capital-intensive manufacturers.

The third signal is margin stability after the easiest comparison effects fade. Profit growth of 437.6% is unlikely to remain a normal benchmark because the prior base included losses and weaker utilization.

The more meaningful test is whether gross and operating margins remain healthy as year-over-year comparisons normalize. Stable margins would suggest technical differentiation and manufacturing discipline.

Falling margins alongside sustained revenue growth would point toward pricing pressure or a capacity race. Falling revenue and margins together would indicate that the cycle supplied more of the first-half lift.

These signals should also be read by industry. Semiconductor foundries require different evidence from software subscriptions, medical equipment, or innovative medicines.

Foundries should show utilization, pricing discipline, and yield. Equipment suppliers should show customer acceptance, installed-base growth, and service demand.

AI hardware vendors should show usable cluster performance and software adoption. Enterprise software businesses should show recurring subscriptions and retention.

The exchange's 2025 review provides a useful historical comparison. It described full utilization at major foundries, strong equipment orders, and early profitability milestones among AI computing companies.

The 2026 results therefore did not emerge from a standing start. They extended trends already visible one year earlier, while producing a much larger aggregate profit outcome.

That continuity supports the structural argument. The scale and concentration of the latest gains preserve the cyclical counterargument.

Readers following an RSSHub 36Kr feed should treat the original alert as the beginning of the analysis, not its conclusion. The early 251-company snapshot correctly identified a dramatic improvement.

The completed market data then showed an even stronger result. What remains unanswered is whether customers, cash flow, and margins will validate it over multiple reporting periods.

Watch the next order updates before accepting claims of permanent change. Then compare those orders with operating cash flow and normalized margins.

If all three remain firm, China's hard-tech companies will have stronger evidence that research, production learning, and supply-chain depth are driving earnings. If they diverge, the industrial cycle will remain the simpler explanation.

The midyear reports have already changed the burden of proof. Skeptics can no longer describe the STAR Market only as a collection of expensive research projects waiting for commercialization.

Supporters, however, must now demonstrate that the reported profits can persist after favorable comparisons and capacity ramps pass. That is the real test behind the headline.