An outdoor-apparel company's chip business quarterly revenue exceeded 179 million yuan for the first time, up 206.92% year-on-year. Toread's Q1 2026 report has forced the market to reassess the substance of its 'outdoor + chip' dual-main strategy.
This growth is not driven by simple chip distribution or low-end packaging, but by a large-model bandwidth compression technology jointly developed with Horizon Robotics. This technology targets one of the AI industry's most pressing bottlenecks: when deploying on edge devices, bandwidth consumed by model parameter movement far exceeds computing power itself.
Technical Breakthrough: How Compression Unlocks Computing Power
Running large models on terminals like phones, cars, and robots has long been hampered by the 'memory wall'—massive intermediate data shuttling between chip and memory consumes bandwidth, throttling computing. Tongtu Semiconductor's three-layer compression system attempts to solve this from three dimensions simultaneously.
The first layer targets input data like video streams, compressing data volume to one-fifth, directly reducing bandwidth pressure from front-end sensors to processors. The second layer targets intermediate data like KV Cache generated during inference, achieving 50% further compression beyond Google's similar results. The third layer targets model parameters themselves, achieving 30% compression for billion-parameter models without retraining.
This means the same automotive-grade chip, after applying compression technology, could support inference tasks two to three times larger. For edge scenarios like autonomous driving and robotics, which are extremely sensitive to real-time performance and power consumption, this directly translates into product competitiveness.
Industry Landing: A Closed Loop from Chip to Smart Equipment
Technical pre-research was declared complete at the 2026 Zhongguancun Forum, with measured verification of breakthroughs in inference performance, bandwidth usage, and business effectiveness. But what truly excites the market is that this technology has entered industrialization—the joint development project with Horizon Robotics is embedding compression technology into its automotive-grade AI chip 'Starry Star 6P' and vehicle operating system 'Kakaxia'.
Horizon Robotics reported 2025 full-year revenue of 3.758 billion yuan, up 57.7% year-on-year, making it a leading domestic pre-installed mass-produced autonomous driving chip company. Deep integration with such a chip platform means Tongtu's compression technology is no longer a lab sample, but a mass-production solution about to be installed in hundreds of thousands of vehicles.
More notably, Toread is building a full-chain autonomous and controllable path from underlying chips to end products. During the 618 shopping festival, its self-developed intelligent exoskeleton Crest C3 started pre-sales, selling over 100 units in a single day on Tmall. Weighing only 1.8 kg with a 10:1 thrust-to-weight ratio, this smart equipment is a direct application scenario for compression technology—exoskeletons must process human posture data in real-time and provide power assistance, demanding extremely high chip power and bandwidth.
Performance Validation: Chip Becomes the First Growth Curve
Toread's Q1 2026 data clearly shows transformation results: chip revenue of 179 million yuan accounted for 36.09% of total revenue. In contrast, traditional outdoor business revenue grew only 8.02% year-on-year, far slower than chips.
Overall, the company reported Q1 revenue of 496 million yuan, up 39.41% year-on-year; net profit attributable to shareholders of 56.4457 million yuan, up 14.44%. Non-GAAP net profit was 63.9395 million yuan, surging 32.36%, indicating improving core business profitability.
At the institutional level, Huaxi Securities maintains a 'buy' rating, forecasting 2025-2027 net profits of 237 million, 369 million, and 454 million yuan respectively. CICC raised its 2026 revenue forecast by 29% to 1.952 billion yuan and EPS forecast by 60% to 0.15 yuan. The logic behind these judgments is precisely the high growth sustainability of the chip business.
Market Outlook: A Trillion-Level Blue Ocean for Edge AI
Tongtu Semiconductor CEO Wang Hongjian predicted at the 2026 Zhongguancun Forum that China's edge AI market will grow from 600 billion yuan in 2026 to 2.1 trillion yuan by 2030. This assessment is not exaggerated—scenarios like autonomous driving, embodied robots, AR/VR wearables, and industrial internet are accelerating, each requiring large models to run on terminal devices.
The value of compression technology is that it doesn't change the model itself, but optimizes its operational efficiency on hardware. This means it can widely penetrate eight major scenarios including personal devices, smart spaces, smart vehicles, industrial internet, embodied robots, and AR/VR wearables, independent of specific chip architectures.
For the textile industry, this technological trend is equally noteworthy. Smart clothing, wearable fabrics, and industrial quality inspection equipment will also require deploying AI models on power- and compute-constrained terminals. Breakthroughs in compression technology may indirectly drive cost reduction and performance improvement of intelligent textile equipment.