Huawei's Next-Gen Agentic AI: Revolutionizing Enterprise Infrastructure at INSPIRE 2026

Huawei has officially unveiled a transformative suite of agentic AI solutions specifically designed for the enterprise sector during the prestigious Huawei Cloud INSPIRE 2026 event. Held at the West Bund International Convention and Exhibition Center in Shanghai, the launch marks a significant leap forward in autonomous digital infrastructure, providing businesses with the tools needed to scale intelligence across their operations.

The new product lineup is engineered to empower the global agentic AI infrastructure. By focusing on efficiency and autonomy, Huawei aims to provide a robust foundation for general and AI compute scheduling. The primary pillars of this release include the Agentic Infra unified infrastructure, a next-generation model training and inference platform, and a comprehensive enterprise-grade agent platform.

Advanced Agentic Infrastructure and AICS Capabilities

The newly introduced Agentic Infrastructure is defined by its "token factory" efficiency, continuous learning capabilities, and secure autonomy. At the heart of this system is AICS, which utilizes an ultra-high bandwidth UnifiedBus (UB) network. This network is capable of supporting massive clusters exceeding 100,000 cards, delivering a staggering total computing power of up to 200 EFLOPS.

Furthermore, AICS dramatically optimizes performance by reducing token generation latency to less than 10 million seconds. With a throughput of 5 million tokens per second across 1,000 cards, the system ensures that AI solutions remain highly responsive, maintaining an online service availability of over 99.5%.

Solving Memory and Scheduling Bottlenecks

To address the common memory bottlenecks associated with complex agents, Huawei introduced the AMS (Agentic Memory Storage) system. By using NPU passthrough to Context Memory Storage (CMS) hardware, it creates a PB-scale memory space. This allows for tiered KV-cache pooling, which not only reduces inference costs but also enables multi-day long-running tasks, fostering better continuous learning for digital agents.

Complementing this is the CCE Volcano unified scheduling engine. This engine optimizes resource utilization by more than 30% through a unique "shared training-inference pooling" method combined with fragmentation consolidation. This ensures that both general-purpose and AI workloads are handled with maximum efficiency.

AgentSphere: A Secure Autonomous Foundation

Security remains a top priority with the launch of AgentSphere. This platform provides a secure and autonomous runtime environment for agents, featuring proactive intent protection. Utilizing ultra-lightweight sandbox technology, AgentSphere can achieve startup times of just 100 milliseconds. It also boasts the capacity to batch-create hundreds of thousands of instances per minute, allowing AI agents to scale securely and rapidly within the cloud.

ModelArts: Next-Gen Training and Inference

The ModelArts platform has been upgraded with four core capabilities designed to streamline the AI lifecycle: Reinforcement Learning as a Service (RLaaS), confidential inference, model routing, and model matrix. Specifically, the MaaS model routing supports three distinct policies: experience-first, efficiency-first, and balanced mode. These policies dynamically route requests to the most suitable model based on specific task characteristics.

Huawei has already deployed over 15 state-of-the-art (SOTA) model services, achieving a scheduling accuracy of over 95%. These advancements have led to an average reduction of 20% in calling costs. The enterprise-level RLaaS allows companies to create complex tasks in just one minute, offering end-to-end visualization and ensuring that large models become smarter and more specific with every use.

What is Agentic AI and how does it differ from standard AI?

Agentic AI refers to systems that can act autonomously to achieve specific goals, rather than just responding to prompts. Huawei's solutions provide the infrastructure for these agents to learn continuously, manage their own memory, and execute complex tasks with minimal human intervention.

What are the main benefits of the AICS network?

AICS provides ultra-high computing power (up to 200 EFLOPS) and extremely low latency. This allows enterprises to process massive amounts of data (up to 5 million tokens per second) while maintaining nearly 100% service uptime, making it ideal for large-scale industrial applications.

How does the AMS system improve AI learning?

The Agentic Memory Storage (AMS) system solves the "memory bottleneck" by creating a PB-scale memory space. This allows AI agents to remember context over long periods and through multi-day tasks, which is essential for deep, continuous learning in enterprise environments.

What is the purpose of the CCE Volcano engine?

CCE Volcano is a scheduling engine that manages how computing resources are used. By pooling training and inference tasks together, it reduces waste and improves overall hardware utilization by more than 30%, lowering the total cost of ownership for businesses.

Is the AgentSphere environment secure for sensitive data?

Yes, AgentSphere is built on ultra-lightweight sandbox technology that provides a secure, isolated environment for AI agents. It includes proactive intent protection to ensure that autonomous actions remain within safe and authorized boundaries.

🔎 In conclusion, Huawei's latest advancements in agentic AI represent a pivotal moment for enterprise digital transformation. By integrating high-performance computing, innovative memory storage, and secure autonomous environments, Huawei is not just providing tools but is building the very foundation of the next industrial era. As these technologies become more accessible through platforms like ModelArts, the potential for businesses to innovate and optimize through AI is virtually limitless.