The robotics sector has seen a significant influx of new participants over the past two years, with over 370 new companies registered nationwide, averaging several new entrants each week. This trend is clearly reflected at the World Artificial Intelligence Conference (WAIC), where the dedicated robot zone has expanded dramatically. From featuring just 25 humanoid robots at its 2024 debut, the zone this year hosts over 200 related companies, growing from a small corner to occupying a substantial portion of the exhibition space. Operating in this highly competitive field, these new players are leveraging distinct "latecomer advantages." At WAIC, four innovative companies with divergent approaches provided their answers, each notably avoiding direct robot "body" manufacturing to instead build their competitive moats from foundational aspects like ecosystems, tactile sensing, data, and models.

Digua Robot made its first appearance at WAIC this year. Founded in 2024 after spinning off from the robotics division of autonomous driving chip giant Horizon Robotics, the team decided from the outset not to build robot bodies. Instead, it chose to be a "tool seller," providing the most effective development tools for all robot creators. After two years of development, Digua's positioning has become clearer: to build the foundational "mother ecosystem" for the robotics era. As explained by Hu Chunxu, Vice President of Developer Ecosystem at Digua Robot, this "mother ecosystem" is not about full-stack in-house development but about creating a standardized, industry-common foundational platform. This platform acts as the "greatest common denominator" between different robot categories, enabling upstream and downstream partners to collaborate, innovate, and connect efficiently on the same platform, collectively driving the industry toward mass production. This strategy stems from an insight into industry trends: future robot categories will be highly specialized, and any breakout product will rely on underlying computing power, development tools, and operating systems. Digua has now accumulated experience in mass production for over a hundred robot models, connected with over 100,000 developers globally, and jointly created millions of robot products, covering the full chain from sensors and actuators to operating systems and higher-level software. This cross-category, differentiated mass-production experience constitutes Digua's most valuable asset.

Xinzhi Jushen (New Wisdom Embodied) exhibits the boldness of a newcomer. Founded just last year, it debuted at WAIC as a "centerpiece exhibit," demonstrating a robotic hand picking up potato chips under the watchful eyes of many established robot "veterans." This year, it again occupied a prime spot, showcasing tactile sensors, tactile data, and tactile models. The company is betting on the embodied intelligence track with a focus on tactile sensing because the team believes that while embodied intelligence is booming, the technological path is far from settled, with various approaches like brain-cerebellum coordination, VLA (Vision-Language-Action) models, and world action models coexisting. Tactile sensing is seen as the crucial "last centimeter" for achieving fine manipulation. Collecting high-quality tactile data demands extremely high equipment specifications: each end-effector requires a tactile sensing module, coupled with force feedback, synchronous data collection, and fine-grained annotation. The collected data includes not just image sequences but also physical information like contact force, slippage, deformation, and texture. Xinzhi Jushen has already built its own professional tactile data collection factory covering over a thousand square meters, handling massive,精细化操作 tasks like industrial assembly, wire harness manipulation, and flexible object handling, using this "data from the fingertips" to build a unique technological barrier. Recently, the company announced the completion of a nearly 100 million yuan angel round of financing, co-led by Shanghai Science and Technology Innovation Group under Shanghai Guosheng Group and Fudan Science and Technology Innovation, with participation from funds including Ceyuan Fund under Shanghai Science and Technology Innovation Group.

Coowa Robot has been running in the robotics race for a full decade. From its inception, Coowa's goal was to create general-purpose AI robots. However, constrained by the AI technology of the time, directly pursuing a general-purpose product was impractical. The team chose to start with the more technologically controllable and demand-rigid scenario of low-speed autonomous driving, diving headfirst into the municipal sanitation sector to first perfect a street-sweeping robot. In 2023, the Transformer architecture saw explosive application-level growth. Seeing the possibility for general-purpose robots, Coowa upgraded its corporate vision to become a "full-stack solution provider for urban embodied intelligence." This February, it publicly released its self-developed Coowa WAM 2.0 general world model foundation, enabling robots to make more intelligent predictions and decisions in complex, dynamic environments. Its core architecture, "one brain, multiple forms," uses a single general AI brain to drive robots of various morphologies, enabling the reuse of cognitive, decision-making, and execution capabilities across different bodies. This is combined with fine-tuning for complex tasks in urban scenarios to expand into broader applications. On the embodied intelligence track, there are many companies like Coowa that first take root in real-world soil, solving practical problems while accumulating core assets, then wait for the technological wave to arrive to make a remarkable leap. After a decade of deep cultivation, Coowa possesses a "gold mine" of urban data: as of this year, its urban scene data exceeds 50 petabytes, with cumulative real-road testing reaching 55 million kilometers, and deployments have reached the scale of tens of thousands of units in over 50 cities and regions.

DaXiao Robot (Big Small Robot), founded just one year ago, is a veritable "latecomer" in the embodied intelligence field. It identified the same technical path as Nvidia, securing two rounds of financing totaling hundreds of millions of US dollars within four months. Its self-developed "Kairos" world model also achieved top rankings across four global embodied intelligence benchmark tests. These four "first-place" achievements correspond to the four core challenges for embodied intelligence落地: authoritative verification of complex manipulation capability, the ultimate test of generalization ability in open scenes, empirical validation of usability in virtual scenes, and industry gold-standard certification for physical modeling capability. In other words, DaXiao leads comprehensively across the three key dimensions of generative, physical, and cognitive intelligence. To understand its "latecomer-first" logic, one must consider industry pain points: the lack of physical logic and causal reasoning, and the困境 of "data scarcity." DaXiao designed the world's first integrated "understanding-generation-prediction" architecture, creating a full closed loop from "knowing the world" to "predicting the world" to "interacting with the world." Simultaneously, it pioneered a human-centric ACE embodied R&D paradigm, capable of accumulating tens of millions of hours of data within a year. Currently, at Shanghai's West Bund, DaXiao's "Xiaotu" robotic dogs have been deployed for常态化 unmanned patrols. In Baoshan's riverside scenic area, it provides an unmanned retail solution for the "SenseMartGo" shaomai (steamed dumpling) shopping robot小店, which is also part of the SenseTime ecosystem, achieving highly efficient operations with records like "fastest order in 15 seconds" and "400 orders per day." This "latecomer" is proving with hard results that it's possible to start later and finish first.