Huawei Cloud Advances Comprehensive Industry AI

Huawei Cloud unveils its “Comprehensive Intelligence” strategy, aiming to be the foundational “black soil” for industry AI. This includes massive compute power, specialized Pangu models for diverse sectors, and embodied AI platforms, empowering enterprises to lead in the intelligent era.

Shanghai, September 19, 2025 – As the digital world converges with an unprecedented wave of artificial intelligence, Huawei Cloud is positioning itself not just as a provider, but as the foundational “black soil” upon which the future of industry intelligence will grow.

On the second day of the tenth HUAWEI CONNECT 2025, Huawei Rotating Chairman and CEO of Huawei Cloud Computing, Zhang Ping’an, laid out a compelling vision: “Comprehensive Intelligence, Becoming an Industry AI Pioneer.”

His address wasn’t merely a technical update; it was a declaration of intent, outlining how Huawei Cloud is meticulously crafting the infrastructure, models, and tools necessary to embed AI deeply into every sector.

The bedrock of this intelligent future, according to Zhang, is raw computing power, and Huawei Cloud is scaling it to monumental levels.

The CloudMatrix384 Ascend AI Cloud Service, now fully operational, represents a significant leap.

But the ambition extends far beyond, with plans to elevate CloudMatrix supernodes from 384 cards to a staggering 8192, capable of forming ultra-large clusters of 500,000 to 1 million cards.

This isn’t just about bigger numbers; it’s about providing an “exuberant” torrent of AI computing power, essential for the increasingly complex demands of the intelligent era.

Innovation extends to memory management too, with the pioneering EMS (Elastic Memory Storage) service designed to expand VRAM with standard memory, dramatically reducing latency in large model multi-turn dialogues and enhancing user experience – a critical detail often overlooked in the race for raw speed.

Behind this digital powerhouse are equally impressive physical foundations.

Huawei Cloud’s deployment of fully liquid-cooled AI data centers in Guizhou, Inner Mongolia, and Anhui highlights a commitment to efficiency and sustainability. Learn more about Huawei’s sustainability initiatives.

These centers boast an impressive 80 kilowatts of heat dissipation per rack and a remarkably low PUE (Power Usage Effectiveness) of 1.1, all managed by AI-driven intelligent operations.

The message is clear: enterprises need not undertake costly and complex data center overhauls.

A simple fiber optic connection can instantly tap into Huawei Cloud’s high-efficiency AI compute and a full suite of dedicated AI cloud services.

Furthermore, the introduction of the CloudMatrix384 AI Token Inference Service simplifies access to this formidable power, masking underlying complexities to deliver direct AI computation results with superior performance, often outperforming competitors like H20 by three to four times in single-card inference scenarios.

But compute power alone is not enough; it must be wielded effectively.

This is where the Pangu large models come into play, steadfastly focused on “solving difficult problems and doing difficult things” for industries.

Huawei’s openPangu initiative is democratizing access to Ascend training and inference, lowering the barrier for developers.

Simultaneously, the company remains dedicated to its closed-source Pangu models, pouring investment into their continuous refinement and expansion.

The goal is to empower clients across diverse sectors to build their own specialized large models, accelerating intelligence from government and finance to manufacturing, healthcare, and even niche areas like coal mining and meteorology, where Pangu models are already demonstrating profound value in over 30 industries and 500 scenarios.

The future of intelligence isn’t confined to virtual realms.

Huawei Cloud is pushing the boundaries into the physical world with CloudRobo, its embodied AI platform.

By offloading complex algorithms and intelligent logic to the cloud, CloudRobo enables robots to be lighter, smarter, and more versatile.

This cloud-centric approach leverages vast computational resources and advanced AI models to break the traditional limitations of robotic capabilities, opening up a myriad of new application scenarios. R2C (Robot to Cloud) protocol is a crucial enabler for this.

The announcement of 20 initial R2C partners signals a burgeoning ecosystem for intelligent robotics.

Underpinning these advancements are Huawei Cloud’s broader strategic initiatives.

The Kunpeng ARM cloud services continue to grow at an astonishing pace, with CPU cores expanding by 67% to 15 million in the past year, supporting over 25,000 applications and solidifying its role in general-purpose computing.

Similarly, the GaussDB database, leveraging supernodes and full pooling, offers an efficient and highly reliable data foundation, shattering traditional architectural limits with simultaneous multi-write/multi-read capabilities across any node, boosting transaction processing by 2.9 times compared to non-supernode deployments.

Finally, ensuring that this intelligence is truly ubiquitous and accessible, Huawei Cloud has engineered a full-scenario distributed cloud solution, ranging from CloudOcean to CloudPond, providing a consistent experience across diverse customer environments.

And to empower the creators of this intelligent future, the enterprise-grade Agent platform Versatile has been launched.

This platform promises an easy-to-use, open environment for developing AI Agents, allowing users to generate sophisticated agents from simple business descriptions and flowcharts in just two steps – a testament to Huawei Cloud’s commitment to making AI development efficient and intuitive.

Zhang Ping’an’s address at HUAWEI CONNECT 2025 painted a vivid picture of a future where AI isn’t just a buzzword, but a deeply integrated, powerful force.

Huawei Cloud isn’t merely offering services; it’s cultivating the very “black soil” where intelligence can flourish, empowering industries to navigate and lead in this transformative era.

The blend of colossal compute power, specialized AI models, embodied intelligence, robust foundational services, and developer-friendly tools forms a comprehensive strategy aimed at making “Comprehensive Intelligence” a tangible reality for every sector.