Edge AI deployment has long been bottlenecked by 'data movement'—during model inference, massive parameters and intermediate data shuttle between chips and memory, making bandwidth and memory the primary constraints on computing power. Toread's subsidiary Shanghai Tongtu Semiconductor and Horizon Robotics recently signed a strategic agreement to tackle this bottleneck with full-stack compression technology. This is not just a technical collaboration between two companies; it marks the transition of large-model compression from the research phase to automotive-grade mass production validation.
Technical Path: Three-Layer Compression Targets Memory Wall
Tongtu has built a three-layer compression system covering input data, intermediate data, and model parameters. For large-capacity inputs like video, it can compress data to one-fifth of its original size, significantly reducing input bandwidth pressure on edge devices. For intermediate data during inference (e.g., KV Cache), Tongtu achieves an additional 50% compression over Google's TurboQuant algorithm, effectively alleviating the 'memory wall' bottleneck. For billion- and trillion-parameter models, it can compress parameters by 30% without retraining, enabling plug-and-play deployment.
This approach aligns closely with Google's TurboQuant algorithm released this year. Wang Hongjian, Vice President of Toread and CEO of Shanghai Tongtu, stated at the 2026 Zhongguancun Forum that 'using compression to free computing power' has become a definitive industry direction. Validation results show breakthroughs in inference performance, bandwidth usage, and business outcomes, suggesting a significant reduction in edge AI deployment costs.
Performance Validation: Chip Business Accelerates
The technical roadmap is backed by financial results. In Q1 2026, Toread reported revenue of 496 million yuan, up 39.41% YoY; net profit attributable to shareholders was 56.45 million yuan, up 14.44%; and non-GAAP net profit reached 63.94 million yuan, up 32.36%. The chip business was the standout—revenue of 179 million yuan, up 206.92% YoY, accounting for 36.09% of total revenue.
The chip business has become the company's primary growth driver, while outdoor business grew only 8.02% YoY, creating a 'chip high-growth, outdoor stable-growth' pattern. Huaxi Securities maintains a 'buy' rating, forecasting net profits of 237, 369, and 454 million yuan for 2025-2027. CICC raised its 2026 revenue forecast by 29% to 1.952 billion yuan and EPS forecast by 60% to 0.15 yuan. Institutional expectations for chip growth are clearly more optimistic than the company's own.
Industry Impact: From Automotive Chips to Full-Chain Autonomy
In this collaboration, Tongtu's full-stack compression technology will be deeply integrated with Horizon's intelligent driving computing platform to optimize inference bandwidth and memory usage on automotive-grade chips. Horizon reported 2025 revenue of 3.758 billion yuan, up 57.7% YoY. In 2026, it launched the cockpit-driving integrated chip 'Starry 6P' and the vehicle intelligent operating system 'Kakaxia', maintaining leadership in intelligent driving computing.
Tongtu's compression technology not only serves as an independent growth driver for the chip business but also provides core technical support for Toread's second-generation intelligent exoskeleton and outdoor smart shoes. During the 618 shopping festival, the self-developed intelligent exoskeleton Crest C3, weighing only 1.8 kg with a thrust-to-weight ratio of 10:1, sold over 100 units in a single day on Tmall. Full-chain autonomy from underlying chips to end products is accelerating.
Market Potential: Edge AI from 600 Billion to 2.1 Trillion
Wang Hongjian predicted at the Zhongguancun Forum that China's edge AI market will grow from 600 billion yuan in 2026 to 2.1 trillion yuan by 2030. Tongtu's full-stack compression can be applied to eight major scenarios, including personal devices, smart spaces, in-vehicle systems, industrial IoT, embodied robots, and AR/VR wearables.
For the textile industry, this trend offers inspiration. Low-cost, low-power edge AI deployment could eventually penetrate smart textile equipment, flexible sensors, and wearable fabrics. When an outdoor apparel company achieves 206% growth through its chip business, traditional textile enterprises should reconsider the integration of technology and textiles.