Huawei Cloud: Pioneering AI and Cloud Computing Innovations for Industries

Huawei Cloud at HUAWEI CONNECT 2025: Innovating AI and Cloud Solutions

On September 21, 2025, during the HUAWEI CONNECT 2025 event, Zhang Ping'an, Executive Director of the Board and CEO of Huawei Cloud, delivered an insightful keynote titled “All Intelligence Empowering AI Pioneers for Industries.” His discourse revolved around groundbreaking innovations and practices that Huawei Cloud has introduced to enhance AI computational services, including foundational AI models, embodied AI, AI agents, and more.

Transforming AI Compute Services with CloudMatrix384

This year marked a significant milestone for Huawei Cloud with the introduction of its AI Compute Service, which is powered by the innovative CloudMatrix384 technology. The specifications of the Huawei CloudMatrix supernode have seen an impressive increase from 384 to an astounding 8,192 cards. This technological advancement enables support for hyperscale clusters operating with 500,000 to 1 million cards, providing robust AI computation resources that are invaluable in today's intelligence-centric era.

Moreover, Huawei Cloud announced a breakthrough with its Elastic Memory Service (EMS), which has achieved a first in the industry by boosting video RAM with additional memory. This novel approach significantly reduces latency in multi-turn conversations using foundational models, substantially enhancing user experience.

Huawei has also established fully liquid-cooled AI data centers in various regions across China including Guizhou, Inner Mongolia, and Anhui. These state-of-the-art facilities support thermal dissipation of 80 kW per rack, dramatically lowering the Power Usage Effectiveness (PUE) to 1.1. Furthermore, they provide AI-based operational management, which negates the need for enterprises to rebuild traditional data centers or construct new ones. Businesses merely need a pair of optical fibers to connect to the center and access effective AI computation and comprehensive cloud AI services via Huawei Cloud.

During his presentation, Zhang highlighted the AI Token service, which abstracts underlying technical complexities and directly delivers the end results of AI computation to users. This allows customers to leverage the inference computation power in the most efficient manner possible. The CloudMatrix384 supernode creates complete pooling of computing, memory, and storage resources, effectively decoupling computing tasks, storage tasks, and AI expert systems. It also transforms serial tasks into distributed parallel tasks, significantly enhancing system performance in terms of inference.

In scenarios involving inference tasks with varied latency requirements—such as online, near-line, and offline—inference, the performance of CloudMatrix384 is found to be three to four times superior per card compared to its predecessor, H20.

Empowering Industries with Tailored AI Solutions

Zhang Ping'an acknowledged Huawei Cloud's commitment to refining its Pangu models by delving into industry-specific scenarios and collaborating closely with clients to address their most pressing challenges. Huawei's openPangu initiative provides best practices for AI training and inference, allowing developers easier access to AI computation power. Notably, Huawei is also developing a closed-source version of the Pangu model and is committed to investing steadily in these models. By understanding client needs through industry-specific scenarios, Huawei aims to assist in developing tailored models to accelerate intelligence transformation across various sectors.

The Pangu models have been successfully applied to over 500 scenarios across more than 30 industries. They have proven pivotal in sectors such as government services, finance, manufacturing, healthcare, coal mining, steel production, railways, autonomous driving, and meteorology.

Cloud Robotics: Shaping Future Intelligence

This year also saw Huawei Cloud unveil the CloudRobo Embodied AI Platform. It leverages sophisticated algorithms and intelligent logic deployed on the cloud to create more agile robots. The vast computation power and advanced AI models available on the cloud empower the robots to overcome the limitations that have traditionally restrained them, expanding their applicability across various scenarios. To facilitate seamless and secure communication between robots and the cloud, Huawei launched the Robot to Cloud (R2C) protocol, with 20 initial partners having already committed to adopting this protocol.

Pulsating Power of Kunpeng Cloud Services

A fundamental strategy for Huawei Cloud is deploying Kunpeng-powered cloud services. These ARM-based services promise performance, security, and reliability. Over the past year, the number of Kunpeng computing cores available on Huawei Cloud has surged from nine million to fifteen million, reflecting a 67% year-over-year increase. Additionally, the Kunpeng platform has been continually optimized for compatibility with common software and has been tailored to support over 25,000 applications, broadening its use case to various computing scenarios beyond transcoding and web applications.

Conclusion: A Future Ready with Cloud Distributed Solutions

Huawei’s GaussDB databases leverage versatile supernode computing to achieve a layered pooling of computation, memory, and storage resources. They allow concurrent read and write operations from any node, moving away from traditional architectures limited to primary node operations alone. Notably, a GaussDB cluster deployed on supernode architecture is capable of processing an impressive 5.4 million transactions per minute, outperforming conventional clusters without supernodes by nearly three times.

As Huawei Cloud continues to innovate with solutions like CloudOcean, CloudSea, CloudLake, and CloudPond, they completely redefine cloud experiences, ensuring a uniformly high-quality interaction regardless of client location. Furthermore, Huawei Cloud's Versatile agent platform facilitates the rapid development of tailored AI agents, enhancing processing efficiency for clients.

With these pioneering initiatives, Huawei Cloud is setting benchmarks in the AI and cloud computing landscape, paving the way for transformative advancements in industry applications and enhancing the capacity for intelligent processing across multiple sectors.