Shanghai, September 19, 2025 /prnewswire/ – Day 2 of Huawei Connect 2025, Champin'An, Executive Director and CEO of Huawei Board of Directors Huawei Cloudgave a keynote speech entitled “All Intelligence: Empowering AI Pioneers for Industries.” He shared Huawei Cloud's Specific innovations and practices such as AI calculation services, basic models, AI, AI agents, and more.
Continuous innovation in AI computing services: Unlocking powerful calculations in an intelligent era
this year, Huawei Cloud We have announced the AI Compute Service equipped with CloudMatrix384. The Huawei CloudMatrix Supernode specifications will be upgraded from 384 cards to 8,192 cards. Supernodes can support hyperscale clusters running on 500,000-1 million cards, thus providing robust AI computing, an invaluable resource in the intelligent era. Huawei Cloud They also announced innovative memory storage with Elastic Memory Service (EMS). This was the first to achieve the industry by expanding video RAM with Memory. This significantly reduces the latency of multi-round conversations in the foundation model, and greatly improves the user experience.
Huawei Cloud We are deploying fully flow-cooled AI data centers China's guizhouInner Mongolia, and Anhui. These AI data centers support 80 kW of thermal dissipation per cabinet, reduce power usage (PUE) to 1.1, and provide AI-enabled O&M. This means that businesses do not need to rebuild traditional data centers or build new data centers. Instead, it requires only a pair of optical fibers to connect to a data center and access efficient AI computing, and a full stack-only AI cloud service Huawei Cloud.
Champin'I pointed it out Huawei Cloud's AI token services abstract the underlying technical complexity and provide users with the ultimate AI computing results directly. This allows users to take advantage of inference computing power in the most efficient way possible. CloudMatrix384 Supernode provides complete pooling of calculations, memory and storage resources, isolates computing, storage and AI expert systems, converts serial tasks into distributed parallel tasks, and greatly improves system inference performance. CloudMatrix384 provides 3-4 times per card average inference performance in scenarios that include inference tasks with various latency requirements such as online, nearline, and offline inference.
At the meeting, Champin“We have announced the official launch of an AI token service powered by CloudMatrix384, which provides customers with outstanding performance, service and quality.
Tackle challenges head on: Help businesses build their own models
Huawei Cloud By diving into industry-specific scenarios, we hone our Pangu model, working with our customers to tackle the most pressing challenges head on, and rethinking what is possible in these industries. Huawei uses OpenPangue to provide best practices for AI training and inference, making it easier for developers to use AI computing power efficiently. Champin“At the same time, it should be noted that Huawei is developing closed source Pangu models. Huawei will continuously increase investment in Pangu models, constantly research industry scenarios to better understand customer requirements, support customers in developing their own industry-specific models, and accelerate intelligent transformation across the industry.
The Pangu model has been applied to over 500 scenarios in over 30 industries. They have played a key role in areas such as government services, finance, manufacturing, healthcare, coal mining, steel, railroads, autonomous driving, and weather.
Move beyond terminals: Allow the evolution of infinite intelligence on the cloud
this year, Huawei Cloud We have launched an AI platform that Cloudrobo has embodied. This allows complex algorithms and intelligent logic to be deployed in the cloud to achieve lighter robots. By leveraging large computing power on the cloud and advanced AI models, the platform makes robot execution more intelligent. Cloud intelligence overcomes the limitations that keep the robot down and allows it to be applied to more scenarios.
To create a unified, open and secure communication channel between the robot and the cloud, Huawei Cloud I've started the Robot to Cloud (R2C) protocol. Zhang Ping'an announced that the first 20 partners of the R2C protocol are officially mounted.
Kunpeng Cloud Services: Enhance industry innovation with software hardware synergies and open ecosystems
One of Huawei Cloud's The key strategy is to develop an ARM cloud service powered by Kunpeng, which provides performance, security and reliability. Over the past year, Kunpeng numbers calculate cores Huawei Cloud It has increased from 9 million to 15 million, up 67%. Additionally, the Kunpeng platform is continuously improved to support compatibility with mainstream software and is adapted to over 25,000 applications. The Kunpeng platform is firmly supported by transcoding, databases, web applications, cloud phones, as well as Kunpeng cloud services that apply to more common scenarios.
GaussDB: Build the foundation for efficient and reliable data based on supernodes and full pooling
Based on the generic computing supernode, Huawei Cloud's The GAUSSDB database provides layered pooling of computing, memory and storage resources, while simultaneously allowing multi-reads and multi-witting on any node, freeing from the limitations of traditional architectures that only the primary node supports data read/write. The GAUSSDB database also supports dynamic load scheduling, significantly improving the performance of concurrent transaction processing. Deployed based on computing supernodes, GAUSSDB clusters can process 5.4 million transactions per minute, resulting in 2.9 times more performance than non-supernode clusters.
All-Scenario Distributed Cloud: Best Computing in Ubiquitous with Local Access
Huawei Cloud We have built a distributed cloud solution that covers all scenarios, including Cloudoceaan, CloudSea, CloudLake, CloudPond. This covers central regions, hotspot areas and edge sites and provides consistency Huawei Cloud Experience wherever your client's business is.
Build an easy-to-use, effective and open platform for developing and running agents
Huawei Cloud We have launched Versatile, an enterprise-grade agent platform. It aims to function as an easy-to-use, effective and open platform for the development and execution of AI agents. With this platform, customers are equipped to quickly develop AI agents that are tailored to their application scenarios.
Based on versatility, users simply need to prepare and enter business documentation and flow charts. After a simple check, the agent can be generated in two steps, greatly improving the efficiency of generation.
In addition to the keynote speech given at the event, Huawei Cloud It brought a variety of agendas, including the Summit Forum and the Round Table. Huawei Cloud It also works with customers and partners to showcase a wide range of innovative technologies and practices in areas such as cloud infrastructure, large-scale models, databases, AI agents, and embodied AI, demonstrating how technology can drive digital and intelligent transformations in the industry.
Photo – https://mma.prnewswire.com/media/2777251/zhang_ping_an_huawei_s_executive_director_board_ceo_huawei_cloud.jpg
