Thursday, October 22, 2026
AI 인프라 · 뉴스 & 분석
홈 › 컴퓨트·클라우드 › 리포트
컴퓨트·클라우드 · 리포트

Chinese AI cloud provider Capital Online (首都在线) reported a book profit in its latest results, with gross margins recovering while cash remains tight and AI compute investment continues.

Margin recovery at a Chinese compute provider suggests pricing is firming, but persistent cash constraints mean its AI capacity expansion still depends on financing.
업계 전문지Slicast · 2026년 10월 21일 13:11 UTC · 중국 · 출처: 钛媒体
중요도 50

Shouhua Online Returns to Book Profit While Continuing Investment in AI Computing Power

On October 21, Shouhua Online (300846.SZ) released its third-quarter 2026 report. For the first three quarters, revenue reached RMB 956 million, up 3.14% year on year. Net profit attributable to shareholders was RMB 8.7962 million, up 108.85%, turning a book profit after earlier losses. Net cash flow from operating activities was RMB 217 million, up 148.37%.

The results are mixed. Net profit attributable to shareholders turned positive for the first time, and operating cash flow was the strongest for the same period in recent years. However, net profit after deducting non-recurring items was still a loss of RMB 56.1803 million. Third-quarter gross margin rose by nearly 9 percentage points from the previous quarter, while single-quarter revenue fell 2.36% year on year. The company's asset-liability ratio climbed to 68.31%. Construction spending continues to consume cash, and unpaid acquisition payments create further funding needs. Profitability is still recovering, and the company faces a new round of capital expenditure. The RMB 1.587 billion private placement, which is still under exchange review, is therefore critically important.

**Profit turnaround has a one-off component, but operations are improving**

In the third quarter alone, revenue was about RMB 290 million, down 2.36% year on year. Net profit attributable to shareholders reached RMB 40.9812 million, offsetting the first-half loss. After deducting non-recurring items, the company still lost RMB 6.3888 million in the quarter.

The turnaround was mainly driven by asset disposal gains and government subsidies. In the first three quarters, the company recorded RMB 55.4260 million in gains from disposing of equipment and other assets, and RMB 17.1945 million in government subsidies classified as non-recurring gains and losses. Including other non-recurring income and expenses, and after deducting income tax and minority interests, the net non-recurring gain attributable to shareholders was RMB 64.9765 million. This amount covered the RMB 56.1803 million loss on a deducted basis, allowing net profit attributable to shareholders to show RMB 8.7962 million.

*(Image source: company third-quarter report)*

On the operating side, lower cost of sales improved gross profit. In the first three quarters, revenue grew slightly, but cost of sales fell from RMB 808 million in the same period last year to RMB 761 million, increasing gross profit by about RMB 76.37 million year on year. Based on the financial statements, the overall gross margin rose from about 12.75% to 20.35%. Based on the interim report, the third-quarter gross margin was about 26.60%, compared with about 11.91% a year earlier.

The interim report's breakdown by business line shows that cloud revenue grew quickly in the first half, while IDC revenue declined. In the first half, cloud host and related services generated RMB 438 million in revenue, up 39.67% year on year, accounting for nearly two-thirds of total revenue. Gross margin was 15.32%, up 10.47 percentage points. Within this, intelligent cloud revenue was RMB 239 million, up 107.76%. Compared with compute cloud, which provides general-purpose computing power, intelligent cloud relies mainly on high-performance computing resources such as GPUs to serve model training and inference needs. IDC services generated RMB 218 million in first-half revenue, down 24.50%, with gross margin stable at about 21.22%, still higher than the cloud business's margin. The rise in cloud revenue and margin was the main source of new gross profit in the first half.

The company says it is reducing homogeneous, low-margin businesses and directing resources toward intelligent computing infrastructure and technical services. It is adding its own computing equipment while also connecting to and managing computing resources owned by other companies, and cooperating with other firms through investments and equity stakes to provide more computing power to customers.

Gross profit has increased but still does not cover operating expenses. In the first three quarters, the company generated gross profit of about RMB 194 million, while combined selling, administrative, R&D and financial expenses totaled about RMB 226 million. A credit impairment loss of RMB 15.4076 million further eroded profit. However, the deducted net loss for the same period narrowed by 51.65% year on year.

**Cash under pressure; private placement mostly funds AI computing power**

Profitability is still recovering, and developing AI computing power requires continued spending on equipment and data centers. In the first three quarters, net operating cash flow was RMB 217 million, up 148.37% year on year. Over the same period, cash paid to acquire fixed assets, intangible assets and other long-term assets reached RMB 280 million, exceeding net operating cash inflow by about RMB 62.87 million. At the end of September, the company's cash and cash equivalents stood at RMB 272 million, down about RMB 106 million from the beginning of the year.

*(Image source: company third-quarter report)*

At the same time, the company's debt levels and interest burden are rising. Based on the third-quarter report, the asset-liability ratio at period end was about 68.31%, nearly 10 percentage points higher than at the beginning of the year. Current liabilities exceeded current assets by about RMB 373 million. Interest expense in the first three quarters rose 30.84% year on year. The company attributes this mainly to new long-term bank loans taken in the second half of 2025 and fixed assets acquired through finance-lease direct leasing.

Part of the new payables consists of unpaid equity acquisition payments. In June, the company announced plans to acquire the remaining 40% stake in Zhongjia Hexin for RMB 236.52 million. Payments will be made in installments, with the controlling shareholder, Qu Ning, providing a guarantee. By the end of September, the company's other payables had risen from less than RMB 50 million at the start of the year to RMB 257 million. The company says this is mainly due to unpaid equity acquisition payments and increased funds borrowed from the controlling shareholder.

According to the interim report, in February the company borrowed a total of RMB 60 million from controlling shareholder Qu Ning. As of the end of June, this amount had not been repaid. It is due in February 2027, and the announced annual interest rate is 4.9%.

In addition, the company plans to continue expanding its computing infrastructure. A revised draft of the prospectus, released on September 22, shows that of the RMB 1.587 billion to be raised, RMB 200 million will be used to repay bank loans, and the remaining approximately RMB 1.387 billion will go to computing infrastructure construction and R&D. Of this, about RMB 787 million will fund phase II of the Huailai computing center, which plans to build cabinets with a total capacity of 40 megawatts for customers to lease for server placement, along with supporting bandwidth and dedicated line services. About RMB 500 million will go to building an intelligent computing center, which will provide computing resources to customers by leasing third-party data center space and purchasing GPU servers and related equipment. About RMB 101 million will be used to develop the Tianshu intelligent cloud platform.

*(Image source: company announcement)*

From a market perspective, as the company shifts toward intelligent computing, demand for computing power is also diverging. A CSC Financial research report notes that inference applications are driving sustained demand for computing power, while high-end resources are in short supply, and some older, mid- to low-end resources face idleness and falling prices. A CICC report on the inference industry chain points out that the gross margin of computing services is shaped by hardware utilization, procurement costs, technical efficiency and pricing power. Scale expansion helps spread fixed costs, but to be profitable over the long term, the company must improve equipment utilization and increase revenue through better services.

On orders, the company's first Huailai project has already found customers. The interim report discloses that the project passed completion acceptance in June and signed a service agreement in July, leasing its entire 10-megawatt cabinet capacity to a large customer. As of September, when the prospectus was signed, phase II had received leasing intentions totaling 52 megawatts of cabinets, and the intelligent computing center had received intended orders for computing services involving 384 high-performance servers. Although the phase II leasing intentions already exceed planned capacity, neither project has yet signed formal agreements.

Once the new data centers and equipment are in use, the upfront construction and procurement costs will be recognized as expenses year by year. The prospectus estimates that once the fundraising projects reach full production, annual depreciation and amortization could increase by up to about RMB 139 million. Equipment prices will also affect profit margins. If server procurement prices rise by 10% with other factors unchanged, the projected average gross margin of the intelligent computing center project over its construction and operating period will fall by 5.45 percentage points. Whether these new costs can ultimately be covered by project revenue will depend on securing orders and on how efficiently the equipment is used.

*Text by Jingxing Zhiyu; author Chen Xinan. First published on the TMTPost app.*

원문 보기
Chinese AI cloud provider Capital Online… · Slicast