Industry | China's Software Sector Grew Amid Divergence in the First Three Quarters, With Deep Adjustments Underway in Industrial Structure

11/16 2026 403

Preface:

China's software industry remains within a growth range, but turning a profit has clearly become much harder.

Divergence is the fingerprint that deep industrial restructuring has left on the data—and the trail left by a shift in value.

Author | Fang Wensan

Image source | Internet

Cutting Open the 11.8 Trillion Yuan Pie

Slicing into the 11.8 trillion yuan pie makes the divergence look even starker. Integrated circuit design revenue grew 23.3% year on year, far ahead of the pack—and still accelerating from 20.1% in the first half.

E-commerce platform technical service revenue grew only 4.9%; software product revenue, 5%; and information security product and service revenue, 5.4%.

Within the same set of statistics, the gap between the fastest and slowest growers was nearly 20 percentage points.

Information technology services accounted for 69.5% of industry-wide revenue—approaching 70%—while the share of software products slid to 20.8%.

Cloud computing and big data services rose from 15.2% of information technology services in the same period last year to 17.9%.

The center of gravity in delivery models has passed a critical point, and the industry's balance is tipping toward services and subscriptions, with AI and data capturing an ever-larger share of value.

Industrial software—a track continuously fanned by policy tailwinds—saw product revenue grow 7.3% year on year in the first three quarters, below the industry-wide average of 8.4% and below the 8.2% of the first half. Policy intensity and growth speed diverged subtly here.

Subsidies, catalogs, and localization targets can generate orders, but they cannot help enterprises absorb delivery costs and price competition. The release of manufacturing IT budgets lagged half a beat behind the pace of document issuance.

The Average Is an Exquisite Illusion

Looking only at the aggregate data of the A-share computer sector would produce an inspiring story.

The 334 constituent stocks posted total operating revenue of about 1,039 billion yuan in the first three quarters, up 12.01% year on year, and total net profit attributable to parent companies of about 31.5 billion yuan, up 34.09% year on year. Double-digit revenue growth and 30% profit growth looked far more respectable than the MIIT figures.

Move the lens one inch closer, and the story falls apart. Among the 334 companies, the top 10 by net profit together earned 32.840 billion yuan—more than the entire sector's 31.543 billion yuan.

In other words, the remaining 324 companies together posted a net loss of 1.297 billion yuan. The number of loss-making companies reached 172, 11 more than in the same period last year. The median revenue growth rate was only 2.74%, and the median growth rate of net profit attributable to parent companies was -10.20%.

The average is held aloft by a few giants; the median is the true temperature of an ordinary software company.

The cash flow data is even less forgiving. In the first three quarters, the sample companies' total net operating cash flow was -43.468 billion yuan, compared with -17.106 billion yuan in the same period last year—the net outflow expanding to 2.5 times.

Payment cycles from government and large enterprise clients are lengthening, and the gap between revenue recognition and collection is being stretched wider and wider. Accounts receivable stood at 366.067 billion yuan, only slightly down from the same period last year; the money sits in others' accounts, while book profit has already been booked into the statements.

There is also one warm number: contract liabilities and advance receipts totaled 229.203 billion yuan, up 47.55% year on year. Orders are locked in the reservoir, and the timing of opening the sluice is determined by the pace of acceptance.

This explains why third-quarter single-quarter revenue grew 16.93% while net profit attributable to parent companies grew only 3.66%: revenue is being recognized faster, while profit is still on the way.

An examination of profit quality cannot be omitted either. Kingsoft Office's net profit attributable to parent companies in the first three quarters was 2.637 billion yuan, up 123.77% year on year, but third-quarter net profit attributable to parent companies fell 72.30% year on year—which the company attributed to fluctuations in investment income from external investment funds.

Hundsun Technologies' net profit grew 26.49%, of which net investment income contributed 212 million yuan. Stripping away the cosmetic effect of investment income, the quality of the core business warrants separate scrutiny.

The most glaring cases come from cross-industry players. Hanbang Gaoke's 2.783 billion yuan GPU contract recognized only 168,100 yuan of revenue in the third-quarter report, and Maixinlin's 267 million yuan deposit corresponded to zero delivery.

As the computing power concept sweeps through the capital market, contract amounts become props for performance, while order conversion rates and cash flow quality are the real divide. In the story of divergence, there is divergence of real gold and silver—and there are also bubbles of paper prosperity.

Between High AI Order Growth and Profit Realization Lies a Chasm

AI is unquestionably the strongest structural variable in the software industry this year, but the first three quarters of data also remind the market that a considerable distance remains between AI revenue growth and AI profitability.

Yonyou Network is a typical sample. In the first three quarters of 2026, the company's operating revenue was 5.756 billion yuan, up 3.1% year on year; within that, cloud service revenue grew 3.9%, subscription revenue grew 14.2%, and BIP revenue grew 9.6%.

In sharp contrast to these single-digit or low-double-digit growth rates, the company's AI-related contract signing amount reached 1.54 billion yuan, up 108.6% year on year, and AI-related revenue was 760 million yuan, up 275.4% year on year. At the same time, the company's net loss attributable to parent companies in the first three quarters still reached 1.247 billion yuan—only narrower year on year.

This set of numbers reveals the real challenge of AI-ification in enterprise software: AI features easily bring new contracts, new modules, and market attention, but they do not mean the old cost structure will immediately disappear.

Model calls, computing resources, data governance, product transformation, and sales delivery still require continuous investment, and legacy software systems must still be maintained. Enterprises are effectively driving the original business while changing the engine.

iFlytek's financial report presents a similar contradiction. In the first three quarters, the company's revenue was 18.187 billion yuan, up 7.05% year on year, and net profit attributable to parent companies turned from loss to profit at about 21.89 million yuan—but it still lost 763 million yuan after deducting non-recurring gains and losses. In the third quarter, the company's R&D investment exceeded 1.7 billion yuan, accounting for more than 25% of quarterly operating revenue.

At the same time, the company actively contracted some smart city business, with G-end revenue down 2.39% year on year, while smart education and smart healthcare revenue both achieved growth of more than 20%.

This shows that within "AI growth" there is also stratification. Revenue driven by large government projects and system integration and revenue driven by sustained use in high-frequency scenarios such as education, healthcare, and office have different economic models.

Capital market samples also show this divergence. Among the computer industry sample compiled by China Merchants Securities, overall revenue grew 12.31% year on year in the first three quarters of 2026, but the median revenue growth rate of individual stocks was only 2.72%; AI computing infrastructure, AI data services, and AI applications were the three sub-sectors with leading revenue growth.

Huatai Securities' statistics on another sample of computer companies showed that the median revenue growth rate in the first three quarters was about 2.74%, the median growth rate of non-recurring net profit was -7.81%, and net operating cash flow fell 14.33% year on year.

A small number of high-prosperity tracks and leading companies are lifting the overall data, while the operating temperature felt by typical enterprises is noticeably lower.

This is also the core of this year's "growth divergence" in the software industry: AI has not pushed the entire industry into high growth at the same time; it is more like a flow of water redistributing demand.

Enterprises that can obtain computing power, data, high-frequency entry points, and industry knowledge get more water; vendors lacking scenarios and able only to follow features may even face the double pressure of increased R&D investment and customer budgets being squeezed out by AI projects.

Conclusion:

Amid industrial expansion, incremental growth is flowing to chips, cloud and data, and scenario-based AI, while low-differentiation software is under pressure. MIIT has proposed to promote intelligent transformation of software enterprises on a large scale by 2028.

The number of integrated models is no longer enough to prove strength; renewal rates, average customer value, inference costs, and delivery efficiency are the tests. Whoever can turn AI from a cost item into measurable value can ride through the growth divergence.

Some materials referenced: Ministry of Industry and Information Technology: "Software Industry Operation in the First Three Quarters of 2026"; Ministry of Industry and Information Technology: "Software Industry Operation from January to June 2026"; The Paper: "China's software business revenue in the first three quarters was 11,803.1 billion yuan, up 8.4% year on year"; China Securities Journal: "'Hard tech' third-quarter reports blossom in many places: computing power chain high growth realized, new tracks gathering strength to set sail"

Solemnly declare: the copyright of this article belongs to the original author. The reprinted article is only for the purpose of spreading more information. If the author's information is marked incorrectly, please contact us immediately to modify or delete it. Thank you.