On the second day of the Huawei Connect 2025, Zhang Ping’an, board member of Huawei and managing director of Huawei Cloud, held a fundamental speech entitled “All Intelligence: Empowering Ai Pioneers for Industries” (All Intelligence: Strengthening AI pioneers for industry). He reported on the innovations and practices of the Huawei Cloud in the areas of AI computing services, founding models, embodied AI, AI agents and much more.
Constant innovation in the field of AI computing services: powerful computing in the age of intelligence
This year Huawei Cloud announced his AI computing service based on CloudMatrix384. The specifications of the Huawei Cloudmatrix Supernodes are expanded from 384 cards to 8,192 tickets. The supernotion can support a hyperscale cluster with 500,000 to 1 million cards and thus offer robust AI calculation performance, an invaluable resource in the age of intelligence. Huawei Cloud also announced an innovative storage solution with his Elastic Memory Service (EMS), which expands the video RAM with memory as a goodness of the industry. This drastically reduced the latency of conversations with several rounds to foundation models and significantly improves user-friendliness.
Huawei Cloud has set up completely fluid-cooled AI data centers in the Chinese regions of Guizhou, Inner Mongolia and Anhui. These AI data centers support heat dissipation of 80 kW per closet, reduce power consumption effectiveness (PUE) to 1.1 and offer AI-based operating and maintenance functions. This means that companies do not have to convert traditional data centers or build new ones. Instead, you only need a few fiber optic cables to connect to the data center and access powerful AI calculation and dedicated full-stack-Ki cloud services on Huawei Cloud.
Zhang Ping’an pointed out that the Huawei Cloud AI token service abstracts the underlying technical complexity and provides users the final AI calculation results. In this way, users can use the computing power of the inferences as efficiently as possible. The Supernode CloudMatrix384 enables the full pooling of computing, storage and storage resources, decoupled arithmetic tasks, memory tasks and AI experts systems and convert serial tasks into distributed parallel tasks, which significantly improves the system’s inference performance. In scenarios with inference tasks with different latency requirements, such as online, nearline and offline inference, CloudMatrix384 provides an average inference performance per card, which is three to four times as high as that of H20.
At the conference, Zhang Ping’an announced the official start of the AI token Service Powered by Cloudmatrix384. The service offers customers superior performance, service and quality.
Taking up challenges offensive: Supporting companies in building their own models
Huawei Cloud has refined his pangu models through the research of industry-specific scenarios and worked with its customers to overcome their most urgent challenges and redefine the opportunities in these industries. Huawei uses OpenPangu to provide best practice for AI training and infection that make it easier to use developers efficiently. Zhang Ping’an noted that Huawei develops the source-open pangu model at the same time. Huawei will continuously increase the investments in Pangu models, examine industry scenarios in order to better understand customer requirements and support customers in developing their own industry-specific models in order to accelerate the intelligent transformation in all industries.
The Pangu models have already been used in more than 500 scenarios in over 30 industries. They have played an important role in areas such as government services, finances, production, healthcare, coal mining, steel, railway, autonomous driving and meteorology.
Beyond terminals: enable unlimited intelligence development in the cloud
This year Huawei Cloud introduced the Cloudrobo Embodied Ai Platform, which uses complex algorithms and intelligent logic in the cloud to implement lighter robots. By using the enormous computing power and the advanced AI models in the cloud, the platform makes the robot version more intelligent. The cloud intelligence overcomes the restrictions that robots have so far hindered and makes them used for more scenarios.
In order to build a uniform, open and safe communication channel between robots and the cloud, Huawei Cloud has introduced the R2C protocol (Robot to Cloud). Zhang Ping’an announced that the first 20 partners of the R2C protocol are officially on board.
Kunpeng Cloud Services: Promotion of industry innovations through software hardware synergies and an open ecosystem
One of Huawei Cloud’s most important strategies is the development of Kunpeng-based ARM cloud services that offer performance, security and reliability. Last year, the number of Kunpeng data cores in the Huawei Cloud rose from 9 million to 15 million, which corresponds to an increase of 67 %. In addition, the Kunpeng platform was continuously improved to support the compatibility with mainstream software and was adapted to more than 25,000 applications. The Kunpeng platform offers solid support for Kunpeng Cloud Services that can also be used for more general computer scenarios in addition to transcoding, databases, web applications and cloud telephones.
Gaussdb: Building an efficient and reliable database based on supernotual and complete pooling
The Gaussdb databases of the Huawei Cloud are based on all-purpose computing sodnodes and enable the layer of pooling of computing, storage and storage resources as well as reading and writing several data on any node, which means that the restrictions of conventional architecture, in which only the primary node, supports the reading and writing of data. GaussDB databases also support dynamic load planning, which significantly improves the performance of parallel transaction processing. A GaussdB cluster based on supernotual can process 5.4 million transactions per minute, which corresponds to a 2.9-fold increase in performance compared to clusters without supernotion.
Distributed cloud for all scenarios: universal and optimal computing power with local access
Huawei Cloud has developed a distributed cloud solution that covers all scenarios, including Cloudocean, Cloudsea, Cloudlake and Cloudpond. This covers central regions, hotspot areas and EDGE locations and ensures a consistent Huawei Cloud experience, wherever the company of the customer is located.
Building a user -friendly, effective and open platform for the development and operation of agents
Huawei Cloud has launched versatile, an agent platform of the company class. It should serve as easy to use, effective and open platform for the development and operation of AI agents. With this platform, customers will be better able to quickly develop AI agents who correspond to their application scenarios.
Based on versatile, users only have to prepare and enter documents for business description and flow diagrams. After simple confirmation, the agent can be generated in two steps, which significantly increases the efficiency of the generation.
In addition to the main lectures at the event, Huawei Cloud offered a number of events, such as: B. summit forums and round table talks. Huawei Cloud also works with customers and partners to present a wide range of innovative technologies and procedures in areas such as cloud infrastructure, large models, databases, AI agents and embodied AI how technology can facilitate the digital and intelligent transformation of industries.
View original content:https://www.prnewswire.com/news-releases/huawei-cloud-forderung-des-fruchtbaren-bodens-fur-datenverarbeitung-starkung-von-ki-pionieren-fur-die-industrie-302562279.html