Acorn Robot
Category: Robotics / Embodied AI
Chinese embodied-AI startup building 'instinct-driven' (zero-data) robotic manipulation using tactile sensing and an on-device decision model, targeting flexible industrial manufacturing. Acorn Robot was founded in 2024. The company is led by Jiang Yao (姜峣). Based in Beijing, China. Team size: 11-50. Total funding raised: 数亿元 RMB (cumulative; seed ~RMB 100M + angel 数亿元). Latest round: Angel. Key investors include China Merchants Capital (招商局创投), NIO Capital (蔚来资本), Shuimu Tsinghua Alumni Seed Fund (水木清华校友种子基金), Puhua Capital (普华资本), Qiantang Materials Lab (钱唐材料实验室).
- Founded
- 2024
- Headquarters
- Beijing, China
- Team size
- 11-50
- Total funding
- 数亿元 RMB (cumulative; seed ~RMB 100M + angel 数亿元)
Value proposition
Acorn Robot's Natus embodied-instinct model gives robots innate, physics-grounded manipulation ability (grip force, slip detection, adaptive contact response) without any pre-training on task-specific data, enabling 'cold-start, plug-and-play' deployment on new production lines in minutes instead of the week-plus required by conventional programmable industrial robots or data-hungry embodied-AI systems.
Products and solutions
Natus AGE-0 (embodied-instinct model driving real-time tactile-motor response), Magis (skill-layer model for continuous refinement via real-world interaction), proprietary 3rd-gen visuo-tactile sensors (self-developed since 2019, 'dynamic slip representation' tech introduced 2020), standardized dual-arm flexible manufacturing units, 'manufacturing-as-a-service' business model.
Unique value
Zero pre-training-data, instinct-driven manipulation stack combining tactile sensing + biologically-inspired reflex modeling, positioned as a 'white-box' alternative to data-driven VLA/embodied AI — deployed live in a global top-tier cosmetics ODM factory within 2 months, reaching commercial revenue and cutting changeover time from ~1 week to 6-40 minutes.
Target customer
Industrial manufacturers needing flexible, high-mix low-volume production (consumer electronics, daily chemicals/cosmetics ODM, food, automotive, new-energy vehicles, biomedical) that require frequent product changeovers.
Industries served
Industrial manufacturing / flexible production (consumer electronics, cosmetics/daily chemicals, food, automotive, new-energy vehicles, biomedical)
Technology advantage
Proprietary 3rd-gen visuo-tactile sensor hardware (7 years iteration, since 2019, claimed ~5 years ahead of industry on 'dynamic slip representation') + neuroscience-grounded instinct model (Natus) requiring no task-specific training data, enabling millisecond-level adaptive response and true plug-and-play deployment; team spans mechanical engineering, neuroscience, and AI with ~15 years combined manipulation research; explicitly rejects VLA (vision-language-action) architecture as fundamentally flawed for real-time contact control.
How they differentiate
Most Chinese embodied-AI/humanoid players (Unitree, AgiBot, Galbot) pursue data-driven, large-scale demonstration/training pipelines (often VLA-based) for general-purpose or humanoid robots; Acorn Robot instead targets a narrower industrial-manipulation niche with a zero-data, neuroscience-inspired 'instinct' approach claiming immediate deployability without data collection, differentiating even from other tactile-sensor players like Pasny Tech by embedding the model directly at the end-effector rather than feeding data into large models.
Main competitors
Unitree Robotics (宇树科技), AgiBot / Zhiyuan Robotics (智元机器人, ~44% global humanoid shipment share in H1 2026), Galbot / Galaxy General (银河通用), Pasny Tech (帕西尼感知科技, tactile-sensing focused humanoid rival)
Key partnerships
Deployed with an unnamed global top-tier cosmetics ODM factory (POC-to-revenue in 2 months), strategic partnerships with leading customers across consumer electronics, daily chemicals, food, automotive and new-energy vehicle sectors (names undisclosed), investor-side ties to China Merchants Capital, NIO Capital.
Notable customers
Unnamed global top-tier cosmetics ODM factory, unnamed strategic customers in consumer electronics, daily chemicals, food, automotive and new-energy vehicle sectors
Major milestones
2018: founder Jiang Yao begins instinct-driven manipulation research post-Harvard, 2019: proprietary visuo-tactile sensor developed, 2020: 'dynamic slip representation' technique introduced, Late 2024: company formally incorporated in Beijing, March 2026: ~RMB 100M Seed round, Aug 10, 2026: Natus AGE-0 model publicly launched alongside Angel round led by China Merchants Capital and NIO Capital.
Growth metrics
Formally founded late 2024 after ~8 years of pre-company R&D (2018-2024); completed two funding rounds within a 4-month span (Seed March 2026, Angel Aug 2026); live industrial deployment reached POC-to-commercial-revenue in 2 months, reducing factory changeover time from ~1 week to 6-40 minutes at a cosmetics ODM client.
Market positioning
Early-stage deep-tech challenger in China's crowded, capital-intensive embodied-AI/humanoid-robotics market (dominated by AgiBot ~44% and Unitree by 2026 shipment share); Acorn differentiates by rejecting the dominant data-driven VLA paradigm in favor of a non-data-driven technical thesis, with a focused industrial-flexible-manufacturing go-to-market rather than humanoid/consumer robots.
Geographic focus
China (Beijing-based), with global industrial customer deployments (cosmetics ODM factory described as serving global markets)
Patents and IP
Proprietary 'dynamic slip representation' tactile technology (first proposed by the team in 2020); specific patent filings not publicly disclosed.
About Jiang Yao (姜峣)
Associate Researcher, Tsinghua University Dept. of Mechanical Engineering (2018-present); Postdoc, Harvard John A. Paulson School of Engineering and Applied Sciences (2017-2018); Postdoc, Tsinghua Dept. of Precision Instrument (2016-2018); PhD, Tsinghua University Mechanical Engineering (2011-2016); BS, Nanjing University of Science and Technology (2007-2011). ~15 years in robotic manipulation research; discovered the 'motor instinct' thesis during Harvard postdoc neuroscience research on human motor learning.
