How AI is Redefining the Capabilities and Boundaries of ICT Infrastructure

• Hyperscalers & Internet Enterprises: Backed by substantial capital, they are engaged in an aggressive capacity expansion, driven by an "arms race" mentality for large-scale model training.
• Mid-to-High-End Enterprises: More pragmatic in their approach, these organizations prioritize a "lease-before-build" strategy—leveraging cloud-based compute to validate business cases before committing to full-scale production.
• Government Clients: Investment is sharply targeted toward specific scenarios, focusing on enhancing public service delivery and governance efficiency.This strategic divergence is directly reflected in H3C’s tiered product portfolio.

This strategic divergence is directly reflected in H3C’s tiered portfolio. To address high-end compute demand, we introduced the H3C SuperPod series. For internet clients, we established a dedicated Internet & Customized Business Group to deliver agile, tailored AI infrastructure. For industry clients, we provide solutions engineered around concrete, high-value application scenarios.

Beneath these strategies lies a fundamental logic: true technological moats are built on sustained, long-term demand. The journey from R&D to commercial deployment is capital-intensive and time-consuming. A single generation of networking or optical silicon requires three to four years of rigorous development—from design to verification. Without a commitment to long-term market demand, even the most advanced breakthroughs risk becoming isolated achievements.

Consequently, H3C positions itself as the Foundation Builder of the AI Leap—an investment philosophy that prioritizes long-term industrial foresight to secure definitive advantages for the next cycle.

Deepening Compute-Network Convergence: Releasing High-Quality, Resilient Service Capabilities

Once the challenge of "compute availability" is addressed, the critical question becomes: Can this power deliver predictable, effective service? The evaluation metrics for infrastructure are shifting from isolated performance indicators to a holistic appraisal of systemic capabilities.

This is most evident in the evolution of network transmission. In the traditional data center era, networks were mere transport conduits; their mandate was simply moving data from A to B. In the AI era, the network has become the nervous system of the AI cluster. It is now a deterministic factor governing overall efficiency. In large-scale training, the execution efficiency of collective communication primitives is inextricably coupled with network states—where congestion, jitter, or load imbalances directly degrade training performance.

Simultaneously, CPUs, GPUs, and network fabrics have entered a phase of cooperative telemetry. Scheduling systems now require real-time awareness of both compute utilization and network equilibrium, necessitating millisecond-level dynamic orchestration. Compute and network are marching toward unprecedented convergence.

The core battlefield has shifted: the true differentiator today is the ability to orchestrate compute, network, and scheduling into a resilient, cohesive system. For mega-scale data centers, stability is the paramount requirement. We see the SuperPod—exemplified by our H3C UniPoD series—as the critical vehicle for this transformation. By leveraging intra-cabinet, fully interconnected communication, it provides the robust architectural backbone required for high-density, high-efficiency AI orchestration.

AI in ALL: Scaling Vertically from Infrastructure Capabilities to Systematic Competencies

The ultimate value of "Computing × Connectivity" must crystallize into high-quality, resilient service capabilities. For H3C, this extends far beyond hardware; it permeates our entire product line, R&D methodology, and technical services. This is the trajectory of our AI in ALL strategy—systematically embedding AI capabilities into our entire portfolio and operational DNA.

This evolution manifests in two critical ways:

• Internal Intelligence: H3C is driving operations and maintenance (Oamp;M) closer to real-world scenarios. By integrating Large Language Models and AI Agents, we have created an intelligent Oamp;M blueprint that enables full-stack visibility, proactive risk prediction, and automated closed-loop optimization.
• Product Intelligence: The fusion of AI and infrastructure is a two-way street. For instance, our AI Firewalls now utilize machine learning for intelligent threat detection, and our CloudOS is evolving to provide unified, intelligent scheduling across heterogeneous compute pools.

As these capabilities mature, AI will extend further down to the device layer, propelling infrastructure from a passive provider of capacity to a proactive system capable of perception, analysis, and collaborative intelligence.

Through this upward expansion, H3C’s dual roles have become sharper:

• As an AI Infrastructure Foundation Builder, we drive continuous improvements in performance and scalability to support the next generation of AI workloads.
• As an AI Application Enabler, we work closely with industries and ecosystem partners to translate technological capabilities into real-world, high-value business outcomes.

This is the definitive path for the landing of AI in ALL, and it remains the primary focal point where H3C will cultivate long-term, pragmatic innovation in the AI era.