Eka Robotics raises $150M to scale vision-force-action models

Securing the capital for scale

Eka Robotics recently announced it has raised $150 million in Series A funding to expand its technological reach. The Cambridge, Massachusetts-based startup intends to use this capital to scale its core technology and increase its customer deployments. This investment marks a significant milestone for the company, which was founded in 2024.

The funding round was spearheaded by Andreessen Horowitz (a16z) and F-Prime, a venture capital group linked to FMR LLC, the parent organization of Fidelity Investments. The round also included continued support from seed investors Obvious Ventures and E14 Fund. Other participants included MFV Partners, Bison Ventures, Glasswing Ventures, Samsung Next, and various Fortune 500 strategic partners.

Investors are looking to improve dexterous manipulation, which refers to the ability of a robot to handle objects with precision and skill. Raghu Raghuram, a general partner at a16z, noted that the company's approach provides the generality and throughput required for actual deployments. He highlighted that the team has the right background to push robotics research while growing commercial operations.

Moving beyond human mimicry

Eka Robotics is building a vision-force-action (VFA) model designed to give robots an intuition of physics. The company argues that for robots to perform complicated work, they must be able to handle changing conditions and deliver useful results every day. Instead of simply mimicking human data, Eka believes robots should master the underlying laws of physics to become truly general-purpose.

By jointly reasoning about vision and forces, the VFA model allows robots to develop an intuition for tasks in the same way humans do. This approach is intended to help robots handle a wide variety of objects that fall outside of their original training data. The goal is to create a system that can adapt to new environments rather than just repeating a set of pre-programmed motions.

The company emphasizes that robots shouldn't just copy what humans do to be effective. They believe that a deep understanding of physics is the key to versatility. This philosophy aims to break the dependence on human demonstrations, which can be limiting when a robot encounters a situation it has never seen before.

Simulation and reinforcement learning

To train its VFA model, Eka utilizes simulation learning to create a high-volume training ground. In a virtual environment, the company can generate thousands of hours of robot experience in just a few minutes. This allows them to vary objects, environments, and physical conditions at a speed that would be impossible in the physical world.

The company asserts that its VFA model is demonstrating that the typical sim-to-real barrier is falling. This barrier refers to the difficulty of transferring skills learned in a computer simulation to a physical robot in the real world. By using reinforcement learning at scale, Eka aims to create a capital-efficient path toward a foundation model for manipulation.

Reinforcement learning is a type of machine learning where an AI learns by trial and error to achieve a specific goal. Eka claims that breadth of experience buys generality, while reinforcement learning at scale buys performance. This combination allows the model to learn complex behaviors quickly and apply them to diverse real-world scenarios.

Real-world performance metrics

The practical application of Eka's technology is centered on high-speed, dexterous manipulation. Robots equipped with this AI can perform roughly 1,500 operations per hour. These operations include complex tasks such as inserting wires or screwing in a light bulb, which require precise coordination of force and movement.

Shyam Kamadolli, a managing director at Fine Structure Ventures, noted that a robot's value on a factory floor depends on its success rate and its speed over thousands of repetitions. He observed that current approaches to generalization are often insufficient for these requirements. He believes Eka's approach is unique because it addresses the hardest problems in robotics by focusing on physics.

By achieving these performance levels, Eka hopes to move robots into roles that require high reliability. The ability to handle complex tasks with high throughput is what makes the technology viable for commercial use. The company's focus on success rates ensures that the robots can deliver useful work every day in demanding industrial settings.

The path to general-purpose robots

The success of the VFA model could signal a shift in how companies approach robotic intelligence. By moving away from human-centric data and toward physics-based reasoning, Eka is attempting to solve the problem of general-purpose manipulation. This could allow robots to be deployed in a wider range of settings without needing to be retrained for every specific object or task.

The funding from a diverse group of investors, including Fortune 500 strategics and major venture firms, suggests high confidence in this direction. These stakeholders are betting that Eka's model will hold up in real-world deployments of complex tasks. The focus remains on creating a scalable and performant way to achieve dexterous robot skills.

As the company scales its technology and customer deployments, the industry will watch how well the VFA model handles increasingly complex environments. The goal is to move from specialized machines to versatile robots that can reason about the physical world. This transition is essential for robots to become truly useful in human environments and industrial workspaces.