Hengwei Technology Acquires Shuheng Information: The First AIRaaS Merger and Acquisition Case in A-Shares, the AI Industry Shifts from "Selling Computing Power" to "Selling Results" | The Frontline
The competitive logic in the AI industry is changing. Recently, Hengwei Technology announced a trading halt to plan the acquisition of 75% of the shares of Shanghai Shuheng Information (hereinafter referred to as "Shuheng Information") through the issuance of shares and cash payment. This merger has become the first case in the A-share market of a listed company acquiring an AIRaaS (Result as a Service) target, marking the transformation of the AI industry from "computing power competition" to "application implementation".
Hengwei Technology's previous AI layout focused on the "hardware end". Public information shows that its core products cover computing network visualization and computing network infrastructure. It has launched AI all-in-one machines based on domestic GPUs such as Ascend, Muxi, and Kunlunxin, as well as hardware solutions such as orthogonal liquid-cooled super nodes, serving operators, intelligent computing centers and other fields.
However, "it is difficult to reach the core business value of customers only by hardware," a person close to Hengwei Technology told 36Kr. Against the background of the accelerating equalization of AI technology, the simple supply of computing power can no longer meet the deep - seated needs of enterprise customers. Customers are more concerned about whether AI can directly solve business problems and bring quantifiable benefits, rather than purchasing a "computing machine".
It is understood that Shuheng Information, founded in 2017, has the core ability to combine large - model technology with industry scenarios. Through its self - developed S - GPTAI engine and Langtree model orchestration platform, Shuheng can quickly achieve a full - process closed - loop from data preparation to deployment and launch, and supports local deployment. At present, its services cover dozens of brands in industries such as fast - moving consumer goods, automobiles, and finance, and its revenue and profits have maintained rapid growth in the past three years.
AIRaaS: Bypass the "selling tools" logic and access the 20% - 60% core profit pool
The core of this merger is that traditional AI service enterprises are trying to solve the long - standing growth "pain points" in the industry. Traditional AI services often stay at the stage of "selling tools". After customers pay to purchase software or hardware, they often find it difficult to see direct business returns due to difficult implementation and unclear effects, and finally adopt a wait - and - see attitude towards AI deployment.
Different from the traditional model of charging by software authorization and human input, AIRaaS takes "quantifiable business results" as the core, requiring service providers to be directly responsible for the final business goals of customers. Only when the agreed results (such as cost - reduction ratio, efficiency improvement) are achieved will they charge, forming an "AI general contractor" model of "providing all materials and labor and guaranteeing results", and even binding risks with customers through refund commitments and effect guarantee clauses.
The commercial value of this model may be reflected in the "breakthrough of the profit pool". Data shows that traditional digital services only cover 1% - 2% of the enterprise's revenue, while AIRaaS can access the core profit pools such as human resources and supply chain, which account for 20% - 60% of the enterprise's revenue, with a value space 20 - 60 times different. For Shuheng Information, it is precisely with this model that it has achieved profitability against the background of widespread losses among AI application enterprises. For hardware manufacturers like Hengwei Technology, it means that they can shift from "low - value - added hardware sales" to "high - margin service revenue".
Industry shift: From "computing power competition" to "scenario implementation"
This merger also reflects the deep - seated trend in the AI industry. With the opening of the source code of leading inference models such as DeepSeekR1, the equalization of AI technology is accelerating, and simple computing power competition is difficult to form a barrier. The industry consensus is that in 2025, the core of AI application competition will shift from "computing power competition" to "scenario implementation" ability, which tests enterprises' understanding of industry needs and the adaptation efficiency between models and business.
From the technical side, after large - model training reaches a certain scale, the performance improvement brought by each unit of computing power input has begun to slow down. From the demand side, the demand for AI in different industries shows refined differences. The medical industry needs AI for accurate diagnosis, the financial industry needs risk assessment, and the education industry needs personalized tutoring. All these require enterprises to break out of the "parameter - only theory" and penetrate into business scenarios.
However, the final effect of this merger still needs time to be verified. At present, the transaction details (such as performance gambling and integration plans) have not been disclosed. Whether the two parties can achieve technical synergy and resource complementarity in the future, and whether they can open up differentiated advantages in the Xinchuang market remain to be seen. But what is certain is that when "paying for results" has become an industry consensus, the focus of industry competition will officially shift from the "computing power base" to "business value".
According to the announcement, Hengwei Technology has been suspended from trading since September 17, and the suspension is expected to last no more than 10 trading days. The subsequent progress of this first AIRaaS merger case in the A - share market may provide a new example for the implementation of the AI industry.