At the recent Huawei Cloud INSPIRE 2026 event in Shanghai, the tech giant unveiled a notable Agentic AI infrastructure, featuring products that promise to reshape enterprise AI solutions. Among the highlights was the introduction of the Agentic Infra, which combines hardware and software to create a new foundation for intelligent systems.
Dr. Peter Zhou, CEO of Huawei Cloud, stated that this initiative marks a shift in computing paradigms, driven by agentic AI technologies. The new infrastructure includes a suite of products designed to enhance efficiency and scalability in AI workloads, addressing the challenges faced by various industries.
Key Innovations in Agentic AI
The centerpiece of Huawei's announcement is the Agentic Infra, which includes several innovative products aimed at optimizing AI operations:
-
AI Cluster Service (AICS): Built on the ultra-high bandwidth UnifiedBus network, this service supports clusters exceeding 100,000 cards. It delivers extraordinary computing power of up to 200 EFLOPS, with token generation latency reduced to under 10 milliseconds, achieving a throughput of 5 million tokens per second across 1,000 cards. Its online service availability reaches an impressive 99.95%, establishing it as an efficient token factory.
-
Agentic Memory Storage (AMS): Utilizing NPU passthrough technology, this solution creates a PB-scale memory space, supporting tiered KV-cache pooling. This advancement reduces inference costs and facilitates multi-day long-running tasks, effectively breaking existing memory bottlenecks and enabling continuous learning for AI agents.
-
CCE VolcanoNext Unified Scheduling Engine: This engine innovates the scheduling of general-purpose and AI workloads through a method called "shared training-inference pooling + fragmentation consolidation." This approach enhances resource utilization by over 30%, significantly improving the efficiency of AI operations.
-
AgentSphere: This product provides a secure and autonomous environment for AI agents, using ultra-lightweight sandbox technology that allows for rapid startup within 100 milliseconds and the capability to batch-create hundreds of thousands of instances per minute, ensuring secure scaling of agents in cloud environments.
-
ModelArtsNext: Designed as a next-generation model platform, it aims to integrate foundation models into enterprise applications, expanding the potential for AI-driven solutions across various sectors.
Industry Zones and Future Directions
In line with its commitment to industry-specific applications, Huawei Cloud announced the establishment of dedicated zones within its Industry AI Foundry, focusing on Smart Healthcare, Embodied AI, Smart Manufacturing, and Scientific Computing. Each zone is tailored to address unique industry challenges, fostering an ecosystem for digital transformation powered by AI.
The introduction of Agentic Infra positions Huawei Cloud as a key player in the AI infrastructure space and sets a high standard for future advancements in enterprise AI solutions. With these tools, businesses can expect enhanced efficiency, reduced costs, and the capability to tackle complex challenges in an increasingly digital world.
Conclusion
Huawei Cloud's latest developments reflect a significant stride towards integrating AI capabilities within industries. The Agentic AI products unveiled are expected to drive innovation and efficiency, paving the way for a new era of intelligent enterprise solutions. As industries continue to evolve, the importance of such infrastructure will become increasingly evident, underscoring the need for businesses to adapt to the changing technological landscape.
The stories that move AI & crypto markets — before the market reacts.
Free. 7am ET. Five stories. 62,400 readers.
