The Zhitong Finance App learned that Huaye Securities released a research report saying that the interim results of Guangdong-Gangwan Intelligent Computing Technology (01396) exceeded expectations, and AI computing power has entered a large-scale cash-out period. The company has now formed a business system with AI computing power technology services as the core, with revenue reaching RMB 2,559 million, an increase of 1020.9%; AI business revenue of RMB 2,383 billion, accounting for 93.1% of total revenue; overall gross margin increased from 16.6% to 22.9%; and profit attributable to shareholders of RMB 275 million. The bank pointed out that the large-scale revenue and profitability of the company's AI business has been further verified.
In the medium to long term, based on the existing AI computing power technology service business, the “quantum power” heterogeneous computing power scheduling platform developed independently by the company, as well as the token-based billing and MaaS (Model-as-a-Service) model, provide incremental options to improve computing power utilization efficiency, expand customer application scenarios, and optimize revenue structures. Heterogeneous scheduling can integrate different chips, different regions, and different types of computing power resources; token-based billing is closer to the actual usage scenarios of large model inference and enterprise AI applications. If the platform's capabilities gradually mature, the company's revenue structure is expected to expand from mainly computing power delivery to scheduling, software, model deployment and industry solution services that are more sticky and have higher gross margin potential.
The core investment themes of Huaye Securities are as follows:
Realization of the scale of performance
The revenue from AI computing power technology services was about 1.52 billion yuan, with a gross profit margin of 35.5%. The large-scale revenue and operating profitability of the company's AI business were verified, and the valuation logic was further switched to AI computing power technology services and platform growth.
Order and production capacity resonate
By the end of August 2026, the intended orders had exceeded RMB 37 billion, of which the total billing scale had been delivered and stabilized; the FP16 had been delivered and operated with a dense computing power of over 50,000 PFLOPS, providing high visibility for future revenue recognition.
Financing and platformization
Financial institutions intend to grant more than RMB 50 billion in credit, and the weighted average cost of additional financing was reduced to 3.5%; the company promoted “quantum power” heterogeneous computing power scheduling, token-based billing, and MaaS to provide options for upgrading from asset-heavy operations to platform services.
Policy Requirements and Domestic Computing Power
In the context of “East Digital and Western Computing”, the industrialization of artificial intelligence, and the policy of autonomous and controllable key computing power, the demand for high-performance, compliant, and locally deployable computing power continues to expand in government enterprises, finance, manufacturing, the Internet and vertical industries.
State-owned Credit and Cooperation Endorsements
The company received more than RMB 1 billion in strategic investment from state-owned institutions such as Shenzhen Futian, CMB International, and Boyue, which helps enhance project credit, financing capabilities and industrial resource collaboration, and supports equipment procurement, computing power cluster construction and continuous business scale expansion.
Technology delivery and operation and maintenance capabilities
The company has full-stack service capabilities from data center planning, computing power equipment deployment, high-speed network construction, to 10,000-card cluster delivery, performance tuning, and continuous operation and maintenance. It has delivered and operated FP16 dense computing power of more than 50,000 PFLOPS. In addition, the company's operation and maintenance failure rate is less than 0.08%, which is significantly superior to industry standards. Large-scale delivery, stable cluster operation, and continuous operation and maintenance capabilities are important guarantees for order conversion, stable billing, and long-term customer cooperation.