⚛️量子位•Stalecollected in 2h
Huawei Cloud Launches New Agentic AI Series
💡Huawei Cloud enters the agentic AI race with new infrastructure tools for enterprise developers.
⚡ 30-Second TL;DR
What Changed
Launch of new Agentic AI product series
Why It Matters
This launch signals Huawei's strategic shift toward supporting autonomous agent development at scale within the enterprise cloud ecosystem.
What To Do Next
Check the Huawei Cloud documentation for the new Agentic AI tools to evaluate their suitability for your enterprise agent orchestration workflows.
Who should care:Enterprise & Security Teams
Key Points
- •Launch of new Agentic AI product series
- •Focus on building infrastructure for AI agents
- •Positioning as the 'silicon black soil' for intelligent applications
🧠 Deep Insight
Web-grounded analysis with 17 cited sources.
🔑 Enhanced Key Takeaways
- •The new Agentic AI series from Huawei Cloud includes an 'Agentic Infra integrated intelligent infrastructure,' a next-generation model training and promotion platform, and an enterprise-level intelligent agent platform.
- •Huawei Cloud's Agentic AI strategy is underpinned by a supernode architecture, specialized industry model training, and an agent platform designed for seamless integration with existing business systems.
- •Agentic AI systems, unlike traditional AI, are engineered for a higher degree of autonomy, enabling them to independently set goals, formulate plans, and execute tasks with minimal human oversight.
- •Huawei's Agentic AI leverages its Pangu models, which are specifically developed for business-to-business (ToB) markets and feature a three-layer architecture comprising foundational large language models (LLMs), industry-specific models, and scenario-specific models.
- •The underlying infrastructure for Huawei's Agentic AI features CloudMatrix384 supernodes interconnected by MatrixLink, optimized to support hybrid computing and scale inference clusters, particularly for Mixture of Experts (MoE) models to enhance NPU throughput.
🛠️ Technical Deep Dive
- Agentic AI systems operate through a continuous cycle of Perception, Reasoning (utilizing Large Language Models), Planning, Action, and Reflection.
- Huawei's infrastructure for Agentic AI incorporates CloudMatrix384 supernodes with MatrixLink interconnects, designed for hybrid compute environments and scaling inference clusters.
- The supernode architecture is specifically optimized for Mixture of Experts (MoE) models, facilitating expert parallelism inference to reduce NPU idle time and potentially increase single-PU inference speed by 4-5 times.
- Huawei's Pangu models employ a three-layer decoupled architecture: L0 for foundational models (e.g., NLP, CV, Multimodal, Prediction, Scientific Computing), L1 for industry-tailored models, and L2 for scenario-specific models.
- The Pangu NLP model 5.5 is a 718B deep thinking MoE model, composed of 256 experts, demonstrating enhancements in knowledge reasoning, tool invocation, and mathematics.
- The Pangu World Model, built upon the Pangu Multimodal Model, is capable of generating digital physical spaces for training intelligent driving systems and embodied AI robots.
- The AgenticCore solution integrates AI into core network functions such as mobile Internet, voice, and operations & maintenance (O&M), evolving Telco Cloud infrastructure to support ubiquitous AI agent access and an agent-based communication network.
🔮 Future ImplicationsAI analysis grounded in cited sources
Agentic AI will significantly automate complex business processes across various industries.
Agentic AI can manage complex tasks, collaborate, and adapt, moving beyond traditional Robotic Process Automation (RPA) to automate entire workflows, as seen in examples like supply chain optimization and customer service.
Huawei Cloud's Agentic AI series will strengthen its position in the enterprise AI market, particularly in China and telecommunications.
Huawei's focus on industry-specific Pangu models, its 'AI in a box' private cloud strategy, and its AgenticCore solution for telecom networks indicate a targeted approach to enterprise and infrastructure clients.
The adoption of agentic AI will drive demand for advanced, specialized AI infrastructure.
Agentic systems demand more from infrastructure than static LLM usage, requiring supernode architectures, memory-centric storage, and optimized interconnects to handle continuous agent operations and large-scale orchestration.
⏳ Timeline
2005
Huawei Cloud division formed.
2011
Huawei established Noah's Ark Laboratory for basic AI research.
2017-09
HUAWEI CLOUD launched the Enterprise Intelligence (EI) platform.
2019
Huawei began building the Pangu large learning model.
2023-07
Huawei launched Pangu 3.0, focusing on enterprise AI with a three-layer architecture.
2025-03
Huawei launched the industry's first AI Core Network at MWC Barcelona 2025.
2025-06
Huawei Cloud announced Pangu Models 5.5 and a new AI Cloud Service, with upgrades across five capabilities and new industry-specific deep thinking models.
2025-10
Huawei Cloud unveiled its supernode architecture and Versatile platform for enterprise agentic AI at Huawei Connect 2025.
2026-03
Huawei unveiled the AgenticCore solution at MWC Barcelona 2026.
2026-06-05
Huawei Cloud officially released its new Agentic AI series, including Agentic Infra integrated intelligent infrastructure, a new generation model training and promotion platform, and an enterprise-level intelligent agent platform.
📎 Sources (17)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
📰
Weekly AI Recap
Read this week's curated digest of top AI events →
👉Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: 量子位 ↗