Eka Robotics has raised US$150 million in a Series A round led by Andreessen Horowitz and F-Prime to build general-purpose robots that perform beyond human capabilities and work safely alongside people. The round also includes participation from Obvious Ventures, E14 Fund, MFV Partners, Bison Ventures, Glasswing Ventures, Samsung Next, and several Fortune 500 strategic investors. The company will use the capital to scale its technology and accelerate customer deployments in dexterous manipulation.
The stakes of dexterous manipulation
Trillions of dollars of value flow through human hands, from packing delicate strawberries to assembling GPU server racks. Bringing robots into this work requires both the versatility to handle changing conditions and the performance to deliver useful work every day. Eka Robotics is starting with dexterous manipulation because it sits at the center of this economic opportunity.
Bridging performance and generality
Modern robotics faces a persistent tension between performance and generality. Commercial deployments demand speed and high accuracy, but most deployed robots have achieved that performance only by sacrificing the ability to handle new tasks and environments. Eka argues that true value requires both, and it is pursuing a path that relies on physics rather than human imitation.
A physics-based approach
The company is building a vision-force-action model, or VFA, to give robots an intuition for physics through large-scale reinforcement learning. Eka says the approach is inspired by systems such as AlphaZero, which achieved superhuman performance in Chess and Go using zero human data. The model jointly reasons about vision and forces so a robot can understand how firmly to grasp a soft object or how to respond when a cable is not fully inserted.
Simulated environments for reinforcement learning
Large-scale reinforcement learning is impractical with real-world data, so Eka generates thousands of hours of robot experience in minutes within simulation. The company varies objects, environments, and physical conditions to teach the model broad, contact-rich manipulation. Eka says the sim-to-real barrier is now falling, making this the most scalable and capital-efficient route to a foundation model for manipulation.
Early results
Eka reports that its robots can pick a wide variety of objects outside the training distribution at roughly 1,500 operations per hour, about two and a half times the human baseline of 600 operations per hour. The same robot can screw in a light bulb and insert wires, pointing to early signs of generalization. An earlier version of the model led WIRED to describe the company as approaching a ChatGPT moment for the physical world.
Investor support
Investors framed the technology as a fundamental shift away from human demonstrations and toward reliable, high-speed physical work. Andreessen Horowitz General Partner Raghu Raghuram said the VFA model is the first approach he has seen that solves manipulation with the generality and throughput needed for real deployments. Other backers pointed to the founding team's technical depth and focus on success rate and speed over thousands of repetitions.
What's next
The new funding will help Eka scale its VFA model to more tasks, expand the capabilities of its robotic systems, and accelerate customer deployments. The company is focused on transforming these capabilities into sustained, reliable performance in real operations across customer sites. Its team includes specialists from institutions such as MIT, Berkeley, DeepMind, Boston Dynamics, Microsoft, Carnegie Mellon, and several other leading research organizations.
Eka Robotics is positioning itself at the intersection of performance and generality for physical work. The Series A round signals growing investor conviction that simulation and reinforcement learning can produce robots capable of superhuman manipulation. As the company scales deployments, it aims to reshape where work happens, how facilities are designed, and how supply chains respond to real-world demand.