Huawei Cloud Drives AI Computing Revolution at HUAWEI CONNECT 2025
At HUAWEI CONNECT 2025, held in Shanghai, a pivotal event for the technology community, Huawei's Executive Director Zhang Ping'an delivered a powerful keynote titled "All Intelligence Driving AI Pioneers for Industries". This annual conference provides a platform for Huawei to demonstrate its commitment to driving innovation and excellence in the realm of cloud computing and artificial intelligence (AI). In his talk, Zhang elaborated on the advancements made by Huawei Cloud in various areas, including AI computing services, foundational models, embedded AI, and intelligent agents.
One of the standout announcements was the introduction of the CloudMatrix384 AI Computing Service. This upgrade from 384 to an astonishing 8,192 nodes allows for a hyper-scale cluster that can handle between 500,000 and 1 million cards, setting a new standard for robust AI computation. This innovation marks a significant step toward establishing powerful computing resources essential for intelligent applications.
In addition to the CloudMatrix384 service, Huawei Cloud unveiled its Elastic Memory Service (EMS) which achieves a remarkable feat by expanding video RAM and substantially decreasing latency for multi-turn conversations in foundational models. This improvement significantly enhances the user experience, ensuring that AI applications run smoothly and efficiently.
The conference also highlighted Huawei Cloud's investment in state-of-the-art data centers equipped with liquid cooling in locations such as Guizhou, Inner Mongolia, and Anhui. These advanced facilities support an impressive heat dissipation of 80 kW per cabinet and optimize energy efficiency with a PUE of 1.1. Businesses now have the flexibility to connect to these AI data centers using just a fiber optic line, simplifying the process and enhancing accessibility to efficient cloud-based AI services without needing to construct their own traditional data centers.
In simplifying the user experience, Zhang spoke about the new AI Token Service, which minimizes technical complexities, allowing users to directly access AI computation results. The CloudMatrix384 supernode efficiently manages resource pooling across computing, memory, and storage, enabling the decoupling of computation tasks from storage tasks. This architecture leads to significant improvements in system inference performance, with the CloudMatrix384 delivering 3 to 4 times better inference performance compared to previous models.
Further emphasizing collaboration and innovation, Huawei Cloud has enhanced its Pangu Models through detailed analysis of specific industry scenarios. By partnering with clients, Huawei aims to tackle the most pressing challenges in various sectors, enhancing their AI capabilities. The Pangu Models, which have been successfully applied across over 500 scenarios in more than 30 industries, demonstrate their effectiveness in diverse fields, including healthcare, finance, and autonomous driving.
On the robotics front, Huawei Cloud introduced the CloudRobo embedded AI platform, which synergizes cloud computing and advanced AI models to create more efficient robots. The new Robot to Cloud (R2C) protocol will enhance communication between robots and the cloud, opening up new avenues for development by integrating the first 20 partners into its ecosystem.
Additionally, Huawei Cloud is paving the way for Kunpeng Cloud Services, focusing on ARM-based cloud offerings. Since last year, computing core numbers have surged from 9 million to 15 million, while the Kunpeng platform has achieved compatibility with over 25,000 applications, supporting various computing scenarios such as transcoding and web applications.
In terms of database management, the GaussDB framework, built on general-purpose supernodes, represents a leap in performance by allowing simultaneous read and write operations across multiple nodes. This drastically increases transaction processing capacity, demonstrating Huawei Cloud's innovations aimed at delivering high-efficiency databases that meet contemporary demands.
Moreover, Huawei Cloud launched a comprehensive distributed cloud solution, covering a broad spectrum of environment scenarios including CloudOcean and CloudLake, ensuring consistent and seamless experiences throughout their clientele's operations.
Lastly, with the introduction of the Versatile platform, Huawei Cloud aims to provide customers with an efficient and open environment for developing and deploying AI agents, streamlining the creation of agents that meet specific application needs. Participants at the conference engaged in forums and discussions to explore the vast array of technologies that Huawei Cloud is pioneering, showcasing its commitment to propelling digital transformation and intelligent solutions across industries.
Overall, HUAWEI CONNECT 2025 served as a significant platform for Huawei to illustrate its technological advancements and AI-driven solutions that are set to redefine industry standards and fulfill the promise of intelligent technology innovations.
One of the standout announcements was the introduction of the CloudMatrix384 AI Computing Service. This upgrade from 384 to an astonishing 8,192 nodes allows for a hyper-scale cluster that can handle between 500,000 and 1 million cards, setting a new standard for robust AI computation. This innovation marks a significant step toward establishing powerful computing resources essential for intelligent applications.
In addition to the CloudMatrix384 service, Huawei Cloud unveiled its Elastic Memory Service (EMS) which achieves a remarkable feat by expanding video RAM and substantially decreasing latency for multi-turn conversations in foundational models. This improvement significantly enhances the user experience, ensuring that AI applications run smoothly and efficiently.
The conference also highlighted Huawei Cloud's investment in state-of-the-art data centers equipped with liquid cooling in locations such as Guizhou, Inner Mongolia, and Anhui. These advanced facilities support an impressive heat dissipation of 80 kW per cabinet and optimize energy efficiency with a PUE of 1.1. Businesses now have the flexibility to connect to these AI data centers using just a fiber optic line, simplifying the process and enhancing accessibility to efficient cloud-based AI services without needing to construct their own traditional data centers.
In simplifying the user experience, Zhang spoke about the new AI Token Service, which minimizes technical complexities, allowing users to directly access AI computation results. The CloudMatrix384 supernode efficiently manages resource pooling across computing, memory, and storage, enabling the decoupling of computation tasks from storage tasks. This architecture leads to significant improvements in system inference performance, with the CloudMatrix384 delivering 3 to 4 times better inference performance compared to previous models.
Further emphasizing collaboration and innovation, Huawei Cloud has enhanced its Pangu Models through detailed analysis of specific industry scenarios. By partnering with clients, Huawei aims to tackle the most pressing challenges in various sectors, enhancing their AI capabilities. The Pangu Models, which have been successfully applied across over 500 scenarios in more than 30 industries, demonstrate their effectiveness in diverse fields, including healthcare, finance, and autonomous driving.
On the robotics front, Huawei Cloud introduced the CloudRobo embedded AI platform, which synergizes cloud computing and advanced AI models to create more efficient robots. The new Robot to Cloud (R2C) protocol will enhance communication between robots and the cloud, opening up new avenues for development by integrating the first 20 partners into its ecosystem.
Additionally, Huawei Cloud is paving the way for Kunpeng Cloud Services, focusing on ARM-based cloud offerings. Since last year, computing core numbers have surged from 9 million to 15 million, while the Kunpeng platform has achieved compatibility with over 25,000 applications, supporting various computing scenarios such as transcoding and web applications.
In terms of database management, the GaussDB framework, built on general-purpose supernodes, represents a leap in performance by allowing simultaneous read and write operations across multiple nodes. This drastically increases transaction processing capacity, demonstrating Huawei Cloud's innovations aimed at delivering high-efficiency databases that meet contemporary demands.
Moreover, Huawei Cloud launched a comprehensive distributed cloud solution, covering a broad spectrum of environment scenarios including CloudOcean and CloudLake, ensuring consistent and seamless experiences throughout their clientele's operations.
Lastly, with the introduction of the Versatile platform, Huawei Cloud aims to provide customers with an efficient and open environment for developing and deploying AI agents, streamlining the creation of agents that meet specific application needs. Participants at the conference engaged in forums and discussions to explore the vast array of technologies that Huawei Cloud is pioneering, showcasing its commitment to propelling digital transformation and intelligent solutions across industries.
Overall, HUAWEI CONNECT 2025 served as a significant platform for Huawei to illustrate its technological advancements and AI-driven solutions that are set to redefine industry standards and fulfill the promise of intelligent technology innovations.
Topics Consumer Technology）
【About Using Articles】
You can freely use the title and article content by linking to the page where the article is posted.
※ Images cannot be used.
【About Links】
Links are free to use.