AI Servers
Huawei's self-operated AI servers, led by the CloudMatrix 384 cluster, deliver high-performance AI computing using Ascend 910C chips and a fully domestic AI infrastructure stack.
CloudMatrix 384 AI Server
Huawei's CloudMatrix 384 is its flagship AI server system, designed to rival Nvidia's GB200 NVL72 while circumventing export restrictions . It integrates 384 Ascend 910C AI chips, compared to Nvidia's 72 GPUs, achieving 300 PetaFLOPS of BF16 compute performance, roughly double that of Nvidia's system . The architecture uses a “supernode” design with high-speed interconnects, enabling the chips to work together efficiently despite individual performance differences . The system also offers superior memory and bandwidth, though it consumes 559 kW per hour, nearly four times Nvidia's power usage .
Self-Operated AI Infrastructure
Huawei's AI servers are part of a vertically integrated ecosystem that includes:
- Ascend AI NPUs for high-performance AI computation .
- Kunpeng ARM-based processors powering servers and AI infrastructure, reducing reliance on x86 CPUs .
- HarmonyOS for AI-native operations, supporting real-time inferencing, intelligent scheduling, and context-aware services .
- Atlas SuperPoD servers and NPU cards for enterprise AI adoption, compatible with PCIe 5.0 upgrades .
- UnifiedBus 2.0 and TaiShan 950 SuperPoD for general-purpose compute workloads with AI inferencing support . This self-operated approach ensures technological sovereignty, allowing Huawei and Chinese enterprises to maintain AI development capabilities despite geopolitical constraints .
AI Software and Ecosystem
Huawei complements its hardware with a full-stack AI software ecosystem:
- MindSpore and CANN frameworks for AI model training and deployment .
- AI Gallery with over 100,000 domain-specific AI assets .
- Cloud-based toolchains for self-service migration, training, and inference, enabling rapid deployment in production environments . The combination of hardware and software allows high resource utilization (up to 90%), elastic scheduling, and rapid provisioning of AI resources in under 30 minutes .
Strategic Implications
Huawei's self-operated AI servers provide a domestically controlled alternative to Western AI infrastructure, addressing supply chain and export restrictions while offering competitive performance for enterprise and research applications . The upcoming Ascend 910D chip is expected to improve power efficiency further, enhancing the sustainability of large-scale AI deployments . In summary, Huawei's self-operated AI servers, anchored by the CloudMatrix 384 and Ascend NPUs, represent a high-performance, fully integrated AI infrastructure that combines hardware, software, and domestic supply chain autonomy to support large-scale AI workloads.
Information GPU Accelerated Cloud Server_ECS_HUAWEI CLOUD
GPU-accelerated cloud server (GACS) provides outstanding floating-point computing power that is great for real-time, highly
Information China''s Z.ai trained a model using only Huawei hardware
Chinese outfit Zhipu AI claims it trained a new model entirely using Huawei hardware, and that it''s the first company to
Information AI-accelerated ECSs_x86 ECS Specifications_ECS Types and
AI-accelerated ECSs, powered by Ascend processors and software stacks, are dedicated to accelerating AI
Information Huawei Cloud Announces Pangu Models 5.5 and All-new AI
Huawei Cloud has enhanced its intelligent assistant, Pangu Doer, leveraging cutting-edge AI compute, advanced
Information OceanStor AI Storage
As large AI models move into the multimodal era, enterprises face challenges in model training and inference, from low training
Information Huawei Launches Open-Access SuperPoD Architecture for All
Yang explained, "Huawei has developed the groundbreaking SuperPoD architecture based on our UnifiedBus
Information All Products and Solutions
Directory of Huawei enterprise IT infrastructure products, solutions, and services.
Information Atlas 500 Pro AI Edge Server User Guide (Model 3000) 18
The Atlas 500 Pro (model 3000) is a 2 U AI edge server powered by Huawei Kunpeng 920 processors, featuring superb computing
Information Huawei releases CloudMatrix 384 supernode: a milestone
Huawei has built a completely independent and controllable AI computing system through full-stack innovation from
Information AI Powers Self-Driving Full-Lifecycle Data Management
Management intelligence → product intelligence: AI powers self-driving full-lifecycle data management. Artificial
Information Huawei launches CloudMatrix 384 server as an alternative to Nvidia''s
CloudMatrix 384 is powered by 384 of Huawei''s most sophisticated AI chips, the Ascend 910C graphics processing
Information Huawei and Cambricon lead China''s AI server chip surge, piling
Huawei and Cambricon are set to dominate China''s AI server market as geopolitics and self-reliance accelerate
Information FusionServer Pro V6 Smart Server | Huawei Enterprise
The intelligent server is equipped with the latest AI training and inference modules to unleash up to 560 TFLOPS
Information Chinese tech firms from Huawei to Inspur push ''all-in
Dozens of Chinese hi-tech manufacturers – from Lenovo Group and Huawei Technologies to
Information Paving the Way for All Intelligence
AI servers, especially AI computing clusters, are different from general-purpose x86 servers
Information Overview
The Atlas 800 training server (model 9010) is an AI training server running on the Huawei Ascend 910 AI Processor and Intel
Information Huawei turns to Arm-based AI servers to survive US restrictions
Genius or fool''s errand? The reason why Huawei is doubling down on the Arm architecture is simple: It has to
Information Huawei Cloud
Huawei Cloud opens up Huawei''s over 30 years of expertise in ICT and digitalization. We dive into digital and provide Everything as
Information Ascend AI Cloud Service | Huawei Cloud
Ascend AI Cloud Service offers instant access to immense yet cost-effective AI computing power, a reliable platform for training and
Information Huawei Launches FusionServer Pro V6 Intelligent Server Based on
Today, the Huawei Computing Product Line officially released the latest member to their family – the next-generation
Information Huawei Unveiled the Latest SuperPoD, Making an AI Infrastructure
For AI computing, the Atlas 950 SuperPoD, powered by UnifiedBus, integrates 64 NPUs per cabinet and can scale up
Information Huawei Connect 2025: Building AI Infrastructure In A Sanctioned World
For AI enablement, Huawei''s open-sourced AI stack with MindSpore, CANN, and openEuler comes together to form
Information AI-Self Security
AI Security Layered Defense As illustrated in Figure 3-1, three layers of defense are needed for deploying AI systems in service
Information Huawei AI Servers Test Effectiveness of US Curbs
Last month, Huawei unveiled a new AI server cluster in China''s Anhui province powered by its in-house Ascend
Information Inside Huawei''s Atlas 960: How China''s SuperPoDs Aim to
Huawei has unveiled what it calls the world''s most powerful SuperPoDs and SuperClusters, marking a decisive step in China''s effort
Information Huawei plans to launch AI PCs with self-developed CPUs, reducing
Huawei, which is under U.S. sanctions, plans to equip its upcoming artificial intelligence (AI) PC products to be
Information Huawei launches new AI server solution – Trivium China
What happened: On April 10, Huawei launched its new AI server system, CloudMatrix 384. Huawei claims it
Information Huawei regains power in Chinese server sector due to AI chips
Huawei has started reclaiming its growth and influence in Chinese server business due to increasing demands for its
Information Huawei''s AI servers show tech advancing despite US curbs
TOKYO -- Huawei Technologies is steadily building up its own artificial intelligence infrastructure with homegrown chips and servers,
Information Huawei launches CloudMatrix 384 server as an alternative to Nvidia''s AI
Huawei launches CloudMatrix 384 server as an alternative to Nvidia''s AI infrastructure stack - SiliconANGLE