Chinese AI Compute Firm Debuts Full Product Suite at Digital Trade Expo
From September 23 to 27, 2026, the 5th Global Digital Trade Expo took place at the Hangzhou Convention and Exhibition Center. Zhongke Tongliang, representing China’s video intelligence and AI computing sector, presented the complete SmarCo product line at the event. This exhibition focuses on the company’s batched product offerings for market release. All products are built on indigenous RISC-V dataflow architecture, eliminating external dependencies; the core SmarCo GC3 chip delivers up to 200 TOPS AI performance with 128GB LPDDR5 ECC unified memory. Products are available for deployment with primary focus on government and enterprise clients.
- Release timeline: Series unveiled during the event
- Core architecture: RISC-V dataflow architecture (domestically developed)
- Key specs: 200 TOPS AI performance, 128-way 1080p@30fps hardware decoding, 128GB unified memory
- Deployment model: Local-first with IT innovation compatibility
- Target scenarios: Smart cities, video AIGC, robotics development, IT innovation storage
Full-Stack Product Portfolio for AI Workflows
SmarCo’s exhibition spans five core product categories across computing, interaction, encoding, analysis, and storage:
Compute Layer: SmarCo AI Workstation targets desktop-scale AI for video analytics, 70B-parameter model local deployment, and robotics. Its single-device capacity replaces traditional servers for large-scale video AI workloads, with hardware decoding support for 128 channels at 1080p@30fps—a figure significantly exceeding consumer-grade devices (typically 1-2 channels), revealing a key surprise: capacity once exclusive to data centers now resides within a desktop form factor.
Interaction Layer: SmarCo AI Digital Employee serves offline service scenarios, integrating digital avatars, enterprise knowledge bases, multimodal interaction, and local computing for four standardized roles: AI lobby manager, training instructor, sales presenter, and meeting assistant.
Video Processing Layer: SmarCo Video Encoding Appliance uses RISC-V chips for chip-level “lossless” intelligent encoding, maintaining existing video system architecture while improving stability in poor network conditions and releasing several-fold storage space without hardware replacement—addressing industry challenges around “can’t transmit, can’t store.”
Video Intelligence Layer: SmarCo Video Intelligent Storage-AI Appliance provides end-cloud-user integrated architecture with extensible algorithm repository. SmarCo Video Analysis Appliance offers “off-the-shelf” local video monitoring including automatic download, frame-by-frame analysis, and live broadcast recording for high-security government and public security clients.
Storage Layer: SmarCo IT-Innovation Enterprise NAS achieves full-stack domestic compatibility with hardware and software. New AI NAS bundles smart search, intelligent agents, and knowledge base capabilities.
Core Technical Comparison
| Product Series | Core Chip | AI Performance | Max Decoding | Memory | Target Use |
|---|---|---|---|---|---|
| SmarCo AI Workstation | SmarCo GC3 | 200 TOPS | 128 x 1080p@30fps | 128GB LPDDR5 ECC | Local model inferencing |
| AI Digital Employee | Not disclosed | Not disclosed | Not disclosed | Not disclosed | Offline service |
| Video Encoding Appliance | RISC-V chip | Not disclosed | Not disclosed | Not disclosed | Video transmission |
| Video Storage-AI Appliance | RISC-V video AI chip | Not disclosed | Not disclosed | Not disclosed | Video analytics |
| IT-Innovation NAS | Not disclosed | Not disclosed | Not disclosed | Not disclosed | Enterprise storage |
Note: Details limited to disclosed information only.
Deployment Recommendations
- Ready-to-deploy cases: AI teams requiring local model inference, large-scale video processing (>100 channels), and high-security compliant deployments for government/state-owned enterprises.
- Consider waiting: Small businesses needing lightweight inference under RMB 100,000 budget—current products target mid-to-large enterprise scale with unspecified entry cost.
Final Thoughts
Zhongke Tongliang’s appearance signals RISC-V’s transition from prototype to规模化 AI computation. Its three-pillar strategy—“video intelligence, storage intelligence, compute localization”—responds to data security concerns while demonstrating国产 chip ecosystems penetrating core AI infrastructure.