Huawei Cloud Unites 20+ AI Firms for Agentic Era

Huawei Cloud has announced a major strategic partnership with over 20 leading artificial intelligence model providers. This collaboration aims to establish a unified commercial ecosystem for the next generation of AI agents.

The initiative was unveiled at the Huawei Cloud INSPIRE Creator Conference on June 5. It marks a significant shift towards integrated infrastructure for autonomous AI systems.

Key Facts from the Launch

• Partnership Scale: Huawei joined forces with 20+ top-tier model vendors, including Zhipu AI, DeepSeek, Minimax, Kimi, StepFun, Baidu, Meituan LongCat, iFlytek Spark, AiSheng Tech, and Shengshu Tech.
• New Paradigm: The company introduced 'Agentic Infra', defined by four core pillars: efficient token factories, continuous learning, integrated general-intelligence scheduling, and secure autonomy.
• Infrastructure Power: The new AICS Lingqu intelligent computing cluster supports up to 100,000 cards with a total computing power of 200 EFLOPS.
• Performance Metrics: Token generation latency is reduced to under 10 milliseconds, with throughput reaching 5 million tokens per second for thousand-card clusters.
• Memory Innovation: The AMS Agentic memory storage solution offers PB-scale capacity using NPU direct access to context memory storage hardware.
• Commercial Goal: The 'Hundred Models, Thousand States' plan seeks to create a systematic business ecosystem that benefits all participants through shared resources and standards.

Defining the Agentic Infrastructure Paradigm

Huawei Cloud is redefining how enterprise AI infrastructure should operate. The newly proposed Agentic Infra paradigm moves beyond simple model hosting. It focuses on the specific needs of autonomous agents that require persistent memory and real-time decision-making capabilities.

This approach integrates four critical components. First, it establishes an efficient token factory to handle massive data processing loads. Second, it enables continuous learning, allowing models to adapt without full retraining cycles. Third, it implements integrated scheduling for both general and specialized intelligence tasks. Finally, it ensures secure autonomy, protecting agent operations from external threats.

The significance of this shift cannot be overstated. Traditional cloud infrastructure was designed for static applications or batch processing. Modern AI agents, however, operate dynamically. They interact with users, other software, and external APIs in real time. This requires a fundamentally different architectural approach.

By standardizing these elements, Huawei aims to lower the barrier to entry for complex AI deployments. Developers will no longer need to build custom infrastructure stacks for each agent application. Instead, they can leverage pre-optimized environments that guarantee performance and security.

This move positions Huawei as a central orchestrator in the Chinese AI market. By bringing together competitors like Baidu and iFlytek under one infrastructural umbrella, they are fostering a collaborative rather than purely competitive environment. This could accelerate innovation across the entire sector.

Technical Breakdown of New Hardware Solutions

The announcement featured the release of four new products under the Agentic Infra banner. The centerpiece is the AICS Lingqu intelligent computing cluster. This system is built on a super-large bandwidth network architecture designed specifically for large-scale model training and inference.

The specifications are impressive. The cluster supports up to 100,000 computing cards. This scale delivers a total computing power of 200 EFLOPS. Such power is essential for training state-of-the-art foundation models that require vast amounts of data and computational resources.

Performance metrics highlight the efficiency gains. Token generation latency has been slashed to under 10 milliseconds. For developers building real-time conversational agents, this low latency is crucial for maintaining natural user interactions. Additionally, the throughput reaches 5 million tokens per second for thousand-card clusters.

Another key component is the AMS Agentic memory storage solution. This addresses a major bottleneck in current AI agent development: context management. Agents often struggle to retain information over long interactions or complex multi-step tasks.

The AMS solution uses NPU direct access to Context Memory Storage (CMS) hardware. This creates a PB-level massive memory space. It supports KV Cache hierarchical pooling, which significantly reduces inference costs. More importantly, it enables day-long long-range tasks, effectively breaking the limitations of traditional short-term memory buffers.

These technical advancements provide the physical backbone for the Agentic Infra concept. Without such high-performance hardware, the software paradigms would remain theoretical. Huawei’s investment here signals a commitment to practical, deployable solutions for enterprise clients.

Industry Context and Competitive Landscape

This announcement arrives at a critical juncture in the global AI race. While Western companies like OpenAI, Microsoft, and NVIDIA dominate much of the conversation, China is rapidly developing its own robust ecosystem. Huawei’s strategy reflects a distinct approach to market consolidation.

In the West, infrastructure providers often compete directly with model developers. In contrast, Huawei is positioning itself as a neutral platform. By partnering with rivals like Baidu and Meituan, they are creating a shared utility layer. This mirrors the early days of cloud computing, where infrastructure became a commodity service.

The focus on Agentic Infra also distinguishes this move from previous generative AI hype. Most current offerings focus on chatbots or content generation. Huawei is betting on the next wave: autonomous agents that can perform complex workflows independently.

This aligns with broader industry trends. Major tech firms are increasingly investing in agentic workflows. However, few have offered a comprehensive infrastructure stack dedicated specifically to this use case. Huawei’s proposal fills this gap by addressing memory, latency, and security holistically.

For international observers, this development highlights the depth of China’s AI ambitions. It is not just about catching up in model quality. It is about building superior operational frameworks for deploying AI at scale. The involvement of 20+ top vendors suggests strong industry buy-in, which could lead to rapid adoption within the domestic market.

What This Means for Developers and Businesses

For enterprises looking to deploy AI, this partnership offers several immediate benefits. The primary advantage is cost reduction through shared infrastructure. By leveraging the AICS cluster, companies can avoid the capital expenditure of building their own supercomputing facilities.

Developers will also benefit from standardized tools. The integration of various models onto a single platform simplifies the development process. Teams can switch between different foundational models without rewriting their entire application stack. This flexibility is vital in a fast-moving market where model capabilities evolve weekly.

The enhanced memory capabilities of the AMS solution open new possibilities for application design. Businesses can now build agents capable of handling complex, multi-day projects. For example, a customer service agent could maintain context across weeks of interaction, providing personalized support that feels genuinely human.

However, businesses must also consider the implications of relying on a centralized ecosystem. While the partnership includes multiple vendors, Huawei controls the underlying infrastructure. This raises questions about vendor lock-in and data sovereignty. Companies should evaluate their long-term strategies before fully committing to this specific stack.

Overall, this initiative lowers the technical barriers to advanced AI deployment. Small and medium-sized enterprises can now access capabilities previously reserved for tech giants. This democratization of AI infrastructure could spur a wave of innovation in sectors ranging from finance to healthcare.

Looking Ahead: Future Implications

The success of this ecosystem depends on execution. Huawei must ensure that the promised performance metrics are consistently delivered across all partner models. Any discrepancy could erode trust among the 20+ participating vendors.

We can expect to see the first wave of applications built on this infrastructure within the next 6 to 12 months. These early adopters will serve as proof points for the viability of the Agentic Infra paradigm. Their success or failure will influence broader market adoption.

Internationally, this move may prompt responses from US-based cloud providers. Amazon Web Services, Microsoft Azure, and Google Cloud may need to enhance their own agent-specific offerings to remain competitive. The global AI infrastructure landscape is becoming increasingly multipolar.

Regulatory scrutiny will also play a role. As AI agents become more autonomous, governments will likely impose stricter guidelines on security and accountability. Huawei’s emphasis on secure autonomy positions them well to meet these future regulatory demands.

Ultimately, this partnership represents a maturation of the AI industry. We are moving from experimental prototypes to robust, scalable commercial systems. The focus is shifting from what AI can do to how reliably it can be deployed in real-world scenarios.

Gogo's Take

• 🔥 Why This Matters: This is not just another cloud update; it is a structural shift towards autonomous agents. By solving the memory and latency bottlenecks, Huawei enables AI that can actually work over long periods, not just chat. This moves AI from novelty to utility for enterprises.
• ⚠️ Limitations & Risks: Centralization carries risks. Relying on a single infrastructure provider for 20+ competing models creates potential vendor lock-in. Additionally, the complexity of managing such a large-scale ecosystem could lead to integration challenges or security vulnerabilities if not managed meticulously.
• 💡 Actionable Advice: Developers should monitor the AMS memory solution closely. If you are building apps requiring long-term context (like legal analysis or project management), test this infrastructure against current alternatives like vector databases. Evaluate if the PB-scale memory offers a tangible cost/performance advantage for your specific use case.