Acorn Robot, a startup focused on embodied manipulation, has closed an angel round of financing co-led by NIO Capital. The company simultaneously unveiled its “Natus AGE-0 Embodied Instinct Model,” a foundation model that it says breaks the industry’s reliance on massive training data.

Carl Guan, a partner at NIO Capital, said the embodied intelligence sector is at a pivotal juncture, moving from research to real-world deployment. “We back pioneers who start from first principles and tackle genuine industrial pain points through technological paradigm innovation,” he said. “Acorn’s instinct-driven approach sidesteps the data arms race, offering low‑energy, highly generalised manipulation that addresses long-standing challenges in flexible manufacturing. The team combines cutting-edge research with engineering discipline, and we strongly endorse their differentiated strategy.”

Guan added that NIO Capital expects Acorn Robot to make autonomous robot operation as ubiquitous and plug‑and‑play as electricity, providing a fundamental infrastructure for global industrial transformation.

Founded in late 2024, Acorn Robot claims to be the world’s first general‑purpose embodied intelligence company built on an “instinct‑driven” paradigm. Its core team, drawn from Tsinghua University and Harvard University, spans mechanical engineering, neuroscience, and artificial intelligence. With nearly 15 years of cumulative experience in robotic manipulation, the team has pursued instinct‑based embodied operation since 2017, completing a full cycle of theoretical innovation, engineering validation, and industrial deployment over nine years.

Breaking the Data Bottleneck

According to Acorn, the global embodied intelligence industry is mired in a debilitating data competition. The prevailing VLA (Vision‑Language‑Action) approach depends on vast amounts of demonstration data, ever‑expanding model parameters, and heavy computational resources. While this method performs well in labs, it falters in real factories—plagued by unstable operation, contact‑force failures, and poor generalisation across scenarios. The industry is trapped in a chicken‑and‑egg dilemma: without real‑world data, models cannot mature; without mature models, real‑world data cannot be reliably collected.

Acorn contends that its newly released Natus AGE‑0 foundation model fills the missing execution layer for robot hardware. By enabling zero‑data cold start, it eliminates the dependency on pre‑training and provides a fresh physical‑interaction framework that adheres directly to real‑world physics.

As the industry’s first tactile‑centric general‑purpose manipulation foundation model, Natus AGE‑0 abandons the dominant data‑pre‑training paradigm altogether.

By leveraging tactile perception and contact‑mechanics principles, it distills universal laws of physical interaction, endowing robots with native manipulation instincts that generalise across different hardware, materials, and tasks without any sample data. It requires no large‑scale annotated scenes, is not bound to specific robot hardware, and is not restricted to particular materials or operating conditions.

Unlike traditional models that memorise and replay fixed trajectories, Natus AGE‑0 mimics human biological behaviour through a self‑adaptive, self‑learning pipeline that progresses from instinctive reflex to behavioural emergence and then to experience reinforcement. Tactile signals serve as the primary reflex stimulus, while an internal behavioural organisation mechanism enables autonomous exploration and flexible interaction under diverse industrial conditions. Following the biological principle of "use it or lose it," the model continuously filters and consolidates reusable operational experiences, allowing for autonomous iterative improvement.

With its native interaction instinct, Acorn's robots can be deployed directly on production lines without prior data accumulation. They generate high‑fidelity, high‑relevance physical interaction data on the fly, creating a virtuous cycle that begins with deployment, proceeds to data capture, and culminates in model refinement. This, Acorn says, marks the transition of embodied intelligence from laboratory demonstrations to large‑scale industrial utility.

Targeting Real Industrial Needs

The embodied intelligence industry is undergoing a fundamental shift in evaluation metrics—competition now hinges on replicable productivity rather than flashy lab demos. In discrete manufacturing, flexible production lines face constant challenges: hardware iterations, frequent product changeovers, material tolerances, and dynamic on‑site disturbances. Traditional automation requires re‑data‑collection and re‑tuning for each new workpiece, often causing weeks of downtime.

Acorn takes a long‑term view, focusing on the hardest flexible‑manufacturing problems and validating its business model through real orders and tangible customer value, rather than chasing short‑term funding narratives or custom integration projects for quick cash.

Jiang Yao, founder of Acorn Robot, explained: “We are not building a ‘specialist athlete’ that outperforms humans in speed, but an ‘all‑round athlete’ that adapts quickly to a wide range of tasks—solving the pain points of flexible manufacturing through extreme generalisation.”

To stay true to its product-first philosophy, Acorn made a deliberate strategic choice: it abandoned high margin custom integration projects to avoid the trap of becoming a generic systems integrator. Instead, it concentrates entirely on standardised product development and commercialisation, pursuing a “purer” path of technology productisation.

Acorn's business strategy is exceptionally rare in the embodied intelligence field, closely paralleling the trajectory of the autonomous driving industry, where companies typically first achieve L2‑level capabilities to get products operational, then accumulate data to make the leap to L3 and L4.

“We prioritise closing the business loop first—making technology valuable enough that customers are willing to pay,” said Jiang. “As our standardised hardware and models are deployed on more production lines, we naturally accumulate the most authentic, relevant, and valuable industrial interaction data, fuelling a sustainable iterative cycle.”

Currently, Acorn is refining standardised dual‑arm flexible production cells and exploring a “Manufacturing‑as‑a‑Service” business model, with the goal of building a distributed discrete manufacturing system. Leveraging its proprietary vision‑tactile sensors, end‑effectors, and the Natus embodied instinct model embedded in the hardware, the company does not sell individual components; instead, it provides customers with one‑stop flexible production services precisely tailored to production scenarios in sectors such as FMCG, daily chemicals, and food—sectors that are characterised by high SKU variety, small batch sizes, and rapid iteration cycles.

In just two months, Acorn completed POC validation on a production line for a leading global cosmetics ODM manufacturer and generated revenue, fully validating its technology‑to‑business closed loop.

With fresh funds, Acorn will accelerate Natus model iteration and invest in product R&D, brand building, and commercialization team expansion. Looking ahead, it aims to become a core disrupter in the trillion‑yuan industrial flexible‑manufacturing market—using embodied instinct technology to lower automation barriers and deliver genuine, scalable productivity beyond the data‑competition deadlock.