Tencent-backed Cloud Leopard Intelligent files for IPO
💡Tencent's largest DPU bet is going public; a key indicator for the future of domestic AI infrastructure hardware.
⚡ 30-Second TL;DR
What Changed
Cloud Leopard Intelligent filed for ChiNext IPO to become the first domestic DPU stock.
Why It Matters
The IPO signals a shift in the Greater Bay Area toward core infrastructure chip design, moving beyond consumer electronics. It highlights the critical role of DPUs in scaling AI data center performance.
What To Do Next
Monitor the performance benchmarks of domestic DPU solutions versus Nvidia's BlueField to evaluate infrastructure cost-optimization strategies.
Key Points
- •Cloud Leopard Intelligent filed for ChiNext IPO to become the first domestic DPU stock.
- •Tencent is the largest shareholder with a 19.7792% stake.
- •The company developed China's first high-performance, programmable DPU SoC with 400Gbps bandwidth.
- •Despite 900% revenue growth in 2025, the company remains in a loss-making phase.
🧠 Deep Insight
AI-generated analysis for this event.
🔑 Enhanced Key Takeaways
- •Cloud Leopard Intelligent (Yunbao Intelligent) was founded in 2021 by former Huawei and Cisco executives, leveraging deep expertise in networking and chip architecture.
- •The company's core product line, the 'Leopard' series, is specifically designed to offload infrastructure tasks like virtualization, storage, and network security from the CPU to improve AI cluster efficiency.
- •The IPO filing reveals that the company has secured strategic partnerships with major Chinese cloud service providers and data center operators to integrate its DPU chips into large-scale AI training clusters.
- •The company's R&D expenditure accounted for over 70% of its total operating expenses in 2025, reflecting the high capital intensity required for advanced SoC development.
- •Cloud Leopard utilizes a proprietary software-defined hardware architecture that allows for field-programmable updates, a critical feature for adapting to rapidly evolving AI networking protocols.
📊 Competitor Analysis▸ Show
| Feature | Cloud Leopard Intelligent | NVIDIA (BlueField-3) | Intel (IPU) |
|---|---|---|---|
| Primary Focus | Domestic AI/Cloud DPU | Global AI/HPC DPU | Data Center Infrastructure |
| Bandwidth | 400Gbps | 400Gbps | 200-400Gbps |
| Programmability | High (Software-Defined) | High (DOCA SDK) | High (P4/OneAPI) |
| Market Position | China-Domestic/Tencent-backed | Global Market Leader | Global Market Leader |
🛠️ Technical Deep Dive
- Architecture: Utilizes a multi-core heterogeneous SoC design integrating high-speed network interfaces with dedicated hardware accelerators for packet processing and cryptographic operations.
- Interconnect: Supports RoCE v2 (RDMA over Converged Ethernet) to minimize latency in distributed AI training environments.
- Programmability: Features a proprietary programmable pipeline that allows users to offload custom network functions without modifying the host CPU kernel.
- Efficiency: Designed to reduce CPU overhead by up to 30% in high-traffic data center environments by handling OVS (Open vSwitch) and storage virtualization tasks natively.
🔮 Future ImplicationsAI analysis grounded in cited sources
⏳ Timeline
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: 36氪 ↗
