When a company known for outdoor gear sees its chip business grow 25 times faster than its outdoor segment, the industry wind has clearly shifted.
In Q1 2026, Toread's chip business generated 179 million yuan in revenue, up 206.92% year-on-year, accounting for 36.09% of total revenue. Meanwhile, the outdoor business grew only 8.02%. The dual-core strategy of 'outdoor + chip' adopted since 2021 is now proving its underlying logic with hard data.
Technical Breakthrough: Tackling the Biggest Bottleneck in Edge AI
Deploying large language models on edge devices has long been constrained by bandwidth and memory. During inference, the massive movement of parameters and intermediate data consumes significant bandwidth, creating the so-called 'memory wall'. The joint development project between Tongtu and Horizon Robotics targets precisely this pain point.
Tongtu has built a three-layer data compression system: for high-capacity input data like video, it compresses to 1/5 of original size; for intermediate data, it achieves an additional 50% compression over Google's results; for models with hundreds of billions of parameters, it compresses by 30% without retraining, with plug-and-play capability. This technology significantly improves inference efficiency on automotive-grade chips.
Horizon Robotics, a leading domestic supplier of automotive AI chips, reported 2025 full-year revenue of 3.758 billion yuan, up 57.7% year-on-year. Its 2026 releases—the cockpit-driving fusion chip 'Starry 6P' and the vehicle OS 'Kakaxia'—have already built the hardware foundation for large-scale edge AI deployment. The partnership essentially binds Tongtu's compression algorithms with Horizon's chip platform to create an integrated software-hardware edge AI solution.
Industrial Deployment: The Critical Leap from R&D to Commercialization
Tongtu's full-stack LLM compression technology had completed R&D and passed real-world testing on key metrics like inference performance and bandwidth usage. The strategic signing with Horizon marks the technology's transition from lab to industrial application.
Wang Hongjian, Toread's VP and CEO of Shanghai Tongtu, noted at the 2026 Zhongguancun Forum that Tongtu's technical path aligns closely with Google's TurboQuant memory compression algorithm, confirming that 'using compression to free computing power' is a definite industry direction. This judgment is backed by clear industry logic: China's edge AI market is expected to grow from 600 billion yuan in 2026 to 2.1 trillion yuan by 2030.
For the textile industry, this trend is not irrelevant. Edge AI's low power and low latency are penetrating industrial IoT and smart wearables. Toread's self-developed smart exoskeleton Crest C3, weighing just 1.8 kg, sold over 100 units on a single day during the 618 shopping festival. The full-chain self-sufficiency from chip to end product is opening new growth avenues for smart outdoor gear.
Performance and Capital Validation
Technical ambition is ultimately validated by performance. In Q1 2026, Toread posted operating revenue of 496 million yuan, up 39.41% year-on-year; net profit attributable to shareholders was 56.45 million yuan, up 14.44%; and non-GAAP net profit was 63.94 million yuan, up 32.36%. The chip business's rapid growth served as the primary growth engine.
Institutional investors have also responded positively. On June 17, Toread hosted 11 institutional investors. Huaxi Securities maintained a 'buy' rating, forecasting net profits of 237/369/454 million yuan for 2025-2027. China International Capital Corporation raised its 2026 revenue forecast by 29% to 1.952 billion yuan and its EPS forecast by 60% to 0.15 yuan. Capital market recognition reflects confidence in the edge AI track and the company's technical roadmap.