SHANGHAI,China,September 19,2025 – Huawei Cloud is accelerating enterprise digital transformation with its advanced AI compute services and the newly launched AI Token Service,powered by the CloudMatrix384 supernode. This technology promises to considerably boost inference performance, offering a 3 to 4 times advancement per card compared to previous models.
Huawei Cloud is enhancing AI capabilities for businesses globally, focusing on innovation in AI compute, foundation models, embodied AI, and AI agents, as detailed by Zhang Ping’an, Huawei’s Executive Director of the Board and CEO of Huawei Cloud, and Charles Yang, Senior Vice President of Huawei and President of Huawei Cloud global marketing and Sales Service, at HUAWEI CONNECT 2025.
- Huawei Cloud launched the AI Token Service powered by CloudMatrix384 for efficient AI computing.
- CloudMatrix384 offers 3-4 times better inference performance per card than H20.
- Customer adoption of Huawei Cloud’s AI Compute Service surged from 321 in 2024 to 1,805 this year.
- Pangu Models are now applied in over 500 scenarios across 30 industries.
- Huawei Cloud provides a global infrastructure with data and AI engines to support enterprises.
AI Token Service Simplifies Computing
Table of Contents
zhang Ping’an explained that the AI Token Service streamlines AI operations by abstracting complex underlying technology.It directly delivers final AI computing results, enabling users to maximize the efficiency of inference computing power. The official launch of this service, driven by CloudMatrix384, was a key declaration at the conference, promising enhanced performance, service, and quality.
Zhang Ping’an, Huawei’s Executive Director of the Board and CEO of Huawei Cloud.
The CloudMatrix384 supernode achieves resource pooling for compute, memory, and storage. It decouples compute, storage, and AI tasks, transforming serial processes into distributed parallel ones. This architecture significantly boosts system inference performance, outperforming H20 by a factor of three to four for each card in online, nearline, and offline inference scenarios.
Investing in AI Growth
Huawei Cloud plans to significantly increase investment in AI and computing infrastructure. The goal is to create a robust surroundings for diverse industry applications. The number of global customers utilizing Huawei Cloud’s AI Compute Service has seen remarkable growth, jumping from 321 in 2024 to 1,805 this year.
Investment in Pangu Models will also continue. The company aims to deeply understand industry needs and help clients develop custom models for faster smart transformation. Pangu models have already made a significant impact, being deployed in more than 500 scenarios across over 30 industries.
Charles Yang, Senior Vice president of Huawei and president of Huawei Cloud Global Marketing and Sales Service.
Empowering Global Enterprises
Charles Yang highlighted the complex challenges businesses face during the digital and intelligent transformation. Huawei Cloud’s core positioning remains consistent, offering a complete support system. This includes a “one global cloud infrastructure and two engines – one with data and one with AI.”
This ecosystem features KooVerse, a global cloud infrastructure providing elastic cloud services. Data enablement services efficiently manage and unlock data value. A one-stop AI advancement platform empowers enterprises to build AI-native applications and innovate more easily. Huawei Cloud aggregates global best practices to aid customer success.
HUAWEI CONNECT 2025, themed “All Intelligence,” is exploring AI from strategic, technological, and ecosystem perspectives. The event, held from September 18 to 20 at the Shanghai World Expo Exhibition & Convention Center and Shanghai Expo Center, showcases new digital and intelligent infrastructure, industry-specific solutions, and development tools. (head elements, dateline, schema, etc.).
