Huawei Computing
ProductsCalculateAscend Computing

Ascend Computing

AI computing products powered by Ascend AI processors, supporting full-scenario AI applications including large model training, inference deployment, and high-performance computing.

Kunpeng 920 General-Purpose Server

Kunpeng 920 General-Purpose Server

TaiShan 200 2280

2U dual-socket rack server based on Kunpeng 920 processor, designed for big data, distributed storage, and cloud computing scenarios.

Kunpeng 920 Self-developed Processor128-core high concurrency capabilityLow-power, energy-efficientFully domesticated ecosystem
Processor:Dual Kunpeng 920
Core Count:Up to 128 cores
memory:32 DDR4-2933 modules
Storage:Supports up to 12 3.5-inch SAS/SATA drives
Atlas 800 Training Server

Atlas 800 Training Server

Atlas 800T A2 Gen

Ascend 91 0B Processor-based 8-card AI Training Server for trillion-parameter large model training, ideal for deep learning and LLM training scenarios.

Ascend 910BLarge Model TrainingHCCS InterconnectionLiquid cooling
AI Processor:Ascend 910B
AI Compute:2000+ TFLOPS FP16
Interconnect:HCCS High-Speed Interconnect
memory:12 8GB HBM
Atlas 800 Inference Server

Atlas 800 Inference Server

Atlas 800I A2 Gen

Equipped with Ascend 310P processors, a high-density AI inference server designed for video analysis, natural language processing, recommendation systems, and other inference scenarios.

Ascend 310PHigh-density inferenceVideo AnalysisLow power
AI Processor:Ascend 310P
AI Compute:1000+ TOPS INT8
memory:6 4GB LPDDR4X
Network:10 0GbE
Atlas 500 Intelligent Edge Node

Atlas 500 Intelligent Edge Node

Atlas 500 A2 Gen

Edge AI inference device powered by Ascend 310 processor, designed for edge computing, intelligent video analytics, industrial quality inspection, and more.

Edge AIFanless designSmart VideoIndustrial Quality Inspection
AI Processor:Ascend 310
AI Compute:22 TOPS INT8
memory:1 6GB LPDDR4X
Network:Gigabit Ethernet
Atlas 300T Training Card

Atlas 300T Training Card

Atlas 300T Pro Gen

PCIe accelerator card based on Ascend 910B, designed for server AI training expansion and supporting large-scale distributed model training.

Ascend 910BPCIe 4.0Large Model TrainingDistributed training
AI Processor:Ascend 910B
AI Compute:320 TFLOPS FP16
memory:6 4GB HBM
API:PCIe 4.0
Atlas 300I Inference Card

Atlas 300I Inference Card

Atlas 300I Duo Gen

PCIe inference card based on Ascend 310 featuring a dual-chip design, suitable for video analytics, image recognition, NLP inference, and more.

Dual-chip designVideo AnalysisImage RecognitionLow power
AI Processor:Ascend 310 x2
AI Compute:44 TOPS INT8
memory:3 2GB LPDDR4X
API:PCIe 3.0

Core Advantages

Kunpeng Ecosystem

Built on Kunpeng processors with full-stack independent innovation for security and reliability

Energy-efficient

Intelligent power management with PUE as low as 1.15 for green, energy-efficient operations.

Reliable and Stable

99.999% uptime, redundant design, automatic failover

Flexible Scalability

Modular design, scalable on demand, supports multiple configurations

Use Cases

Covers all scenarios including core government and enterprise business, AI computing, cloud computing, and big data storage.

Core Business for Government and Enterprises

High-performance computing platform for critical business systems such as ERP, CRM, and databases.

AI Training and Inference

Deep learning model training, large-scale inference clusters, and LLM deployment

Cloud Computing and Virtualization

Huawei Cloud Stack, Virtualization Platform, Private Cloud, and Hybrid Cloud Infrastructure

Big Data Storage

Distributed storage, data lake, cold data archival, video surveillance storage

Industry Solutions

Deeply rooted in industry scenarios, providing customized computing solutions for government, finance, telecom operators, and manufacturing sectors.

Government & Public Services

Government Cloud, Smart City, Public Safety

Financial Services

Core Banking, Securities Trading, and Risk Management Systems

Carrier

5G Core, Edge Computing, Network Cloudification

Smart Manufacturing

Industrial Internet, Digital Twin, Intelligent Quality Inspection

Product Documentation

Technical Specifications and User Manual

View more →

Firmware Download

iBMC Firmware and BIOS Update

View more →

Technical Support

One-on-One Expert Support

Contact Us →

Contact Huawei to get a professional computing solution

Our expert team provides one-on-one consulting to design a server configuration tailored to your needs.

Beijing ICP License No. 2026026760