Deploying large models on edge devices has long been bottlenecked by the physical challenge of data movement. During model inference, the transfer of massive parameters and intermediate-state data consumes substantial bandwidth, preventing full utilization of computing power. The recent partnership between Toread's (300005.SZ) subsidiary Shanghai Tongtu and Horizon Robotics directly targets this critical pain point, aiming to elevate edge AI deployment efficiency by an order of magnitude through full-stack large-model bandwidth compression. This collaboration signals that the industry's 'memory wall' problem is moving from labs toward industrialized solutions.
Technical Pathway: Three-Layer Compression Targets Computing Bottlenecks
Tongtu's technological moat covers the entire chain from input to output. The first layer compresses high-volume input data such as video to one-fifth of its original size, significantly reducing input bandwidth pressure on edge devices. The second layer targets intermediate data (e.g., KV Cache) during inference, achieving an additional 50% compression beyond Google's TurboQuant algorithm, effectively alleviating memory wall bottlenecks. The third layer compresses model parameters for billion- and trillion-parameter models by 30%, all without retraining and with plug-and-play capability.
The industry significance of these figures is clear: as edge AI chip computing power growth slows, compression technology becomes the most direct path to unlocking existing potential. Wang Hongjian, Toread's Vice President and CEO of Shanghai Tongtu, noted at the 2026 Zhongguancun Forum that Tongtu's technical route aligns closely with Google's, confirming that 'liberating computing power through compression' is an industry-certain direction. For procurement parties and solution integrators, this means the same chip can handle larger models and more complex tasks, directly reducing hardware upgrade costs.
Performance Validation: Chip Business Becomes Primary Growth Driver
The confidence behind the technology is ultimately validated by financial data. In Q1 2026, Toread reported revenue of 496 million yuan, up 39.41% year-on-year, and net profit attributable to shareholders of 56.45 million yuan, up 14.44%. The chip business stood out with revenue of 179 million yuan, a 206.92% year-on-year surge, accounting for 36.09% of total revenue, surpassing the outdoor business (8.02% growth) to become the company's primary growth driver.
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. CICC raised its 2026 revenue forecast by 29% to 1.952 billion yuan and its EPS forecast by 60% to 0.15 yuan. For investors and procurement parties in the chip supply chain, these figures indicate that Toread's chip business has moved from 'concept validation' into 'volume scaling,' with its technology gaining dual recognition from both market and capital.
Industry Impact: Full-Chain Self-Sufficiency from Chips to End Products
Since establishing its 'outdoor + chip' dual-business strategy in 2021, Toread has acquired five chip-related companies, building a full-chain chip product matrix covering perception, interaction, and display. The partnership with Horizon represents the first large-scale industrialization of Tongtu's full-stack compression technology in automotive-grade chips. Horizon, a leading domestic automotive AI chip provider, posted 2025 revenue of 3.758 billion yuan, up 57.7% year-on-year, with its 'Starry 6P' chip and 'Kakaxia' OS forming a complete intelligent driving computing platform.
Tongtu's compression technology not only serves as an independent growth pole but also provides core technical support for Toread's second-generation intelligent exoskeleton and smart outdoor shoes. During the 618 shopping festival, Toread's self-developed intelligent exoskeleton Crest C3 launched pre-sales, achieving over 100 units sold in a single day on Tmall, with a weight of just 1.8 kg and a thrust-to-weight ratio of 10:1. Full-chain self-sufficiency from underlying chips to end products is accelerating.
Wang Hongjian predicted at the forum that edge AI will form a vast blue ocean market, with China's edge AI market size expected to grow from 600 billion yuan in 2026 to 2.1 trillion yuan by 2030. Leveraging its full-stack compression capability, Tongtu can penetrate eight major scenarios including personal devices, smart spaces, smart vehicles, industrial internet, embodied robots, and AR/VR wearables. For the textile industry's smart transformation, this technical path offers a reference point—the combination of industrial internet and embodied robots is reshaping traditional factory production logic.