As large language models move from the cloud to end-side devices, the computing bottleneck is no longer the chip itself but data movement. In Q1 2026, Toread's subsidiary Shanghai Tongtu Semiconductor signed a strategic agreement with Horizon Robotics to jointly develop LLM bandwidth compression technology. This move directly targets the most challenging 'memory wall' issue in end-side AI deployment—during model inference, the massive transfer of parameters and intermediate data consumes substantial bandwidth, becoming a key bottleneck limiting computing power release.
Technology Breakthrough: Three-Layer Full-Stack Compression
Tongtu has built a compression system covering three layers: input data, intermediate data, and model parameters. Specifically, large-volume input data like video can be compressed to 1/5 of its original size, significantly reducing input bandwidth pressure on end-side devices. For intermediate data in LLMs, compression can achieve an additional 50% reduction beyond Google's results, effectively alleviating the memory wall bottleneck during inference. For trillion-parameter models, parameters can be compressed by 30%, without retraining and with plug-and-play functionality.
At the 2026 Zhongguancun Forum, Toread's VP and Tongtu CEO Wang Hongjian noted that this technology aligns closely with Google's TurboQuant memory compression algorithm in core pain points, technical paths, and industry direction, jointly confirming that 'using compression to free computing power' is an industry certainty. This means Tongtu's technical route is not isolated but resonates with global top-tier solutions, reducing the risk of industrial implementation.
Performance Validation: Chip Business as First Growth Curve
The confidence behind the technology is ultimately validated by performance data. Toread's Q1 2026 report shows revenue of 496 million RMB, up 39.41% year-on-year; net profit attributable to shareholders of 56.45 million RMB, up 14.44%; and non-GAAP net profit of 63.94 million RMB, up 32.36%. Notably, the chip business recorded revenue of 179 million RMB, a 206.92% year-on-year increase, accounting for 36.09% of total revenue.
The chip business has become the company's primary growth driver, while the outdoor business grew 8.02%, forming a robust 'chip high-growth, outdoor stable-growth' pattern. Institutional signals are also positive. On June 17, Toread received 11 institutional investors. Huaxi Securities maintains a 'buy' rating, forecasting net profits of 237/369/454 million RMB for 2025-2027. CICC raised its 2026 revenue forecast by 29% to 1.952 billion RMB and EPS forecast by 60% to 0.15 RMB.
Strategic Depth: Dual-Business Synergy Accelerates Full-Chain Autonomy
Since establishing its 'outdoor + chip' dual-core strategy in 2021, Toread has acquired five chip-related companies, building a chip product matrix covering the full chain from 'perception to interaction to display.' Tongtu's full-stack compression technology serves both as an independent chip business growth pole and as core technical support for Toread's second-generation smart sports exoskeleton and outdoor hiking smart shoes.
During the 618 shopping festival, the company's self-developed smart sports exoskeleton Crest C3 began pre-sales, achieving daily sales of over 100 units on Tmall. The device weighs only 1.8 kg with a thrust-to-weight ratio of 10:1. Full-chain self-control from underlying chips to terminal products is accelerating. Wang Hongjian predicted at the Zhongguancun Forum that China's end-side AI market could grow from 600 billion RMB in 2026 to 2.1 trillion RMB by 2030. Tongtu's full-stack compression capability can be applied to eight major scenarios, including personal devices, smart spaces, smart vehicles, industrial internet, embodied robots, and AR/VR wearables.