AI All-in-One Machine Partner Salon for Power Generation Industry Held in Shanghai

The Anchor of Ecosystem: Not the Sole Protagonist, but a Connector

Liu Jia, Vice President of Huawei China, outlined Huawei’s positioning through four “joint initiatives”: jointly build an open technical base to serve as public infrastructure for computing power; jointly develop industry-level AI scenario maps, enabling real customer value to drive large-scale industrial implementation; jointly establish knowledge and model sharing mechanisms, allowing industrial expertise to be continuously accumulated and self-optimized via algorithms; jointly construct a secure and trustworthy intelligent system to safeguard the bottom line of operational decision-making.

“Huawei does not aim to be the sole protagonist of the ecosystem. Instead, we act as an accelerator for partner growth, a fair referee for benefit distribution, and a connector for intelligent transformation across industries.” This statement set the tone for the entire salon.

The Edge Computing Distinction: The Inflection Point from “Studying” to “Working”

Wang Qingwen, President of Huawei Edge Computing All-in-One Machine Product Line, shared incisive observations. Huawei has launched the edge computing all-in-one machine product line, targeting the inflection point where inference is truly integrated into production systems.

“If model training is like going to school, inference is like going to work. Last year we were still verifying the outcomes of ‘studying’. With the rise of Agents this year, AI is ready to start ‘working’.” He defined the all-in-one machine clearly: it is not merely a server, but an AI subsystem that independently executes tasks within production systems — which may take the form of a single card or four servers. Huawei adheres to open-source principles; its inference acceleration engine and CANN operators are open source for partners to conduct independent optimization.

“If tokens equate to money, the tokens you save through optimization translate directly into revenue generated in production.” He further encouraged partners to develop Agents and Skills. “In the AI era, competition lies not only in chips but also in power supply. If AI can boost power generation efficiency and produce more electricity, this positive cycle can be sustained.”

Strategic Commitment: Bringing Intelligent Computing to Every Power Plant

Li Jiguang, General Manager of Huawei China Government & Enterprise Power System Department, reaffirmed Huawei’s firm commitment: “Huawei’s AI strategy is to bring intelligent computing to every enterprise, every household and every individual.”

He noted that not all enterprises need to build clusters with tens of thousands or hundreds of thousands of cards. Many scenarios only require a single server or even one computing card. While today a single server can run one intelligent agent, future solutions may adopt dual-server and other configurations to orchestrate multiple agents simultaneously. He emphasized the importance of platform development, where partners can openly share challenges and policy needs to accelerate the rollout of all-in-one machines.

Formula for Hit Products: Three Core Questions and Four Key Implementation Factors

Li Shihao, Head of Huawei Computing Power AI All-in-One Machine Business, condensed the methodology into three questions for successful products: Is the scenario valuable and a genuine business need? Can customers afford it? Can partners make profits?

Centered on four pillars — valuable scenarios, extreme cost performance, commercial sales and out-of-box usability — Huawei has jointly developed 10 all-in-one machines with partners and achieved robust growth.

From scenario selection and solution design, verification and development, product pricing to joint marketing, Huawei stands ready to accompany partners through the full product lifecycle.

Benchmark Proof: Algorithms Translate into Tangible Coal-Conservation Benefits

During the roundtable discussion, real-world cases illustrated the above viewpoints. Xin Chunjun from Shudian Yunwang shared the combustion optimization all-in-one machine deployed for 4×650,000 kW units of Datang Zhejiang. Manual operations on the boiler side were eliminated entirely, coal consumption for power supply dropped by 2.2 grams per kWh, yielding annual energy-saving and carbon-reduction benefits of two to three million yuan.

He Lirong of Beijing Biket Technology presented a performance-guaranteed project at Wushashan Power Plant. With the deployment of intelligent algorithms and verified by Xi’an Thermal Power Research Institute, coal consumption per kWh was reduced by 2.073 grams, and 38 manual operations were fully automated.

Li Yuzhen from Shanghai Changgeng introduced the intelligent monitoring dashboard project at Jingran Thermal Power Plant, which achieved remarkable results. Zhao Zigang of Beijing Yueneng Technology developed a new energy daily report intelligent agent, cutting engineers’ workload by 70%–80%. The promise of “delivering value right out of the box” has become reality.

Gao Shengzhe of Xi’an Shengtong Technology presented two product lines: Shengtong Intelligent Inspection and Shengtong Intelligent Query. He also mentioned that the company still seeks improved computing power when adapting its CV Vision Transformer models on Huawei 310 chips, proving that ecosystem co-construction remains an ongoing journey.

The discussions also uncovered new challenges: out-of-box usability and out-of-box value are two separate matters. Without reforming production relations, new quality productive forces will eventually be hindered by legacy mechanisms. Bringing intelligent computing to on-site operations and returning value to customers represents the shared consensus between Huawei and its partners, as well as the next stage of intelligent transformation for the power generation industry.