On the second day of HUAWEI CONNECT 2025, Zhang Ping’an, Executive Director of the Board of Huawei and CEO of Huawei Cloud, delivered a keynote titled “All Intelligence: Empowering AI Pioneers for Industries.” He highlighted Huawei Cloud’s latest advancements in AI Compute Services, core models, embodied AI, AI agents, and more, showcasing the company’s commitment to driving intelligent transformation across industries.
Constant Innovation in AI Compute Service
Zhang emphasized that Huawei Cloud’s AI Compute Service, powered by CloudMatrix384, represents a significant leap in AI computing. This year, Huawei upgraded CloudMatrix supernodes from 384 to 8,192 cards, supporting hyperscale clusters with 500,000 to 1 million cards. These innovations provide robust AI compute, a critical asset in today’s intelligent era. Furthermore, Huawei introduced the Elastic Memory Service (EMS), which combines video RAM with memory, drastically reducing latency for multi-round AI model calls and enhancing user experience.
AI Authority Trend: Huawei Unveils Upgraded Xinghe AI Network Security Solution
Huawei Cloud has also established fully liquid-cooled AI data centers in Guizhou, Inner Mongolia, and Anhui, supporting 80 kW heat dissipation per cabinet and reducing PUE to 1.1. With AI-powered O&M, companies can now connect via a few optical fibers without renovating or building new data centers, gaining access to efficient AI computing and full-stack cloud services.
Zhang noted, “Huawei Cloud’s AI Token Service eliminates the underlying technical complexity and directly provides users with the final AI computational results.” By decoupling compute, storage, and AI expertise systems, the CloudMatrix384 supernode converts serial tasks into distributed parallel ones, delivering 3–4 times faster inference performance than previous systems.
Empowering Industry-Specific AI Models
Huawei Cloud continues refining Pangu Models, tailoring them to industry-specific scenarios and collaborating with customers to solve their most pressing challenges. Zhang highlighted that Huawei provides openPangu for best practices in AI training and inference while also developing closed-source Pangu Models. These models now power over 500 scenarios across 30+ industries, including government, finance, healthcare, autonomous driving, and meteorology.
Advancing Embodied AI and Cloud Intelligence
This year, Huawei launched the CloudRobo Embodied AI Platform, enabling lighter, more intelligent robots through cloud-based algorithms. The R2C (Robot to Cloud) Protocol establishes secure communication between robots and the cloud, with the first 20 partners already onboard.
AI Authority Trend: Huawei Unveils Upgraded Xinghe AI Fabric solution for Latin American Enterprises
Kunpeng Cloud Services and GaussDB
Huawei Cloud’s Kunpeng-powered ARM services expanded compute cores by 67% to 15 million, supporting 25,000+ applications. Meanwhile, GaussDB databases utilize supernodes and resource pooling, achieving 5.4 million transactions per minute—nearly 3x the performance of non-supernode clusters.
Versatile AI Agent Platform and Distributed Cloud
Huawei also introduced Versatile, an enterprise-grade AI agent platform, enabling quick development with minimal setup. Its distributed cloud solution, including CloudOcean, CloudSea, CloudLake, and CloudPond, ensures consistent computing power for all users, regardless of location.
In addition to keynote sessions, Huawei Cloud hosted forums, roundtables, and live demos, reinforcing its leadership in AI, cloud computing, and digital transformation across industries.
AI Authority Trend: Huawei Unveils AI Data Center Solution for Intelligent Computing Era
To share your insights, please write to us at sudipto@intentamplify.com
