Eka Robotics Raises $150M Series A for Force-Aware Robots

A robot purchase can become a long negotiation with the workplace around it. Eka Robotics wants more of that adjustment to happen inside the machine, using force-aware learning to help robots handle changing physical conditions.

The Cambridge, Massachusetts company announced a $150M Series A on October 20, 2026, led by Andreessen Horowitz and F-Prime. Co-founders Pulkit Agrawal and Tuomas Haarnoja are developing a Vision-Force-Action model for robotic manipulation, with the financing intended to expand its capabilities and accelerate customer deployments.

For manufacturers and automation operators, the proposition reaches beyond a faster demonstration. A machine that carries useful physical skills into a different task could change how much engineering a buyer must commission each time the work changes. Eka's opportunity sits inside that purchasing decision, where the cost of introducing a robot includes the effort needed to make it useful.

What Eka Robotics' Series A finances

Eka's new capital backs a company pursuing dexterous manipulation: the physical work of holding, moving and fitting objects as their position or behavior changes. The company identifies broader model capability and customer deployment as its next priorities. Those plans connect the research agenda to a commercial obligation, because a customer needs the capability to survive the circumstances of its own workplace.

The investor group includes returning seed leads Obvious Ventures and E14 Fund, alongside MFV Partners, Bison Ventures, Glasswing Ventures, Samsung Next and unnamed Fortune 500 strategic investors. The announcement identifies F-Prime as affiliated with FMR, Fidelity Investments' parent. That relationship should not be counted as a separate, additional Fidelity investment.

The $150M figure describes this Series A. The primary announcement does not disclose the earlier seed amount or a verified cumulative funding total, so adding an estimated seed figure would give readers an accounting answer the company has not supplied. Likewise, a participating strategic investor is not automatically a customer, a distribution agreement or a purchase commitment.

Those distinctions matter to the deployment story. Funding gives Eka resources for development and commercialization; the source record leaves customer identities, contract economics and the pace of rollout open. An operator evaluating the technology still has to understand the proposed task, the environment and the responsibilities attached to an installation.

Why force feedback changes the deployment conversation

Eka's Vision-Force-Action approach combines visual perception with information about physical contact. Andreessen Horowitz's investment note describes the model, sensing, controller and hardware as a joint design. That organization makes contact part of the robot's decision process while a movement is happening.

A connector illustrates the distinction. Knowing where a socket sits helps a machine approach it; resistance after contact supplies another kind of information about how the insertion is proceeding. The commercial ambition is to make adjustment part of the machine's learned behavior, reducing the need to engineer away every variation in advance.

For the buyer, that would change the boundary of the project. A purchasing team could ask which variations the supplier's existing skills can absorb, which require additional learning and which still call for changes to the work area. Each answer affects the engineering effort that follows the equipment purchase. The buyer can then price a particular task and the work required to keep it running.

That lens also makes reliability easier to discuss without turning it into a slogan. Repeating a successful action, recovering from an awkward contact and maintaining the required work rate belong in the same customer conversation. A capability demonstrated under one set of conditions gives the parties a starting point for defining what the installation must accomplish elsewhere.

The founders' research meets a customer's operating day

Pulkit Agrawal's MIT profile identifies him as an associate professor and Eka co-founder, with research spanning perception, control, hardware and learning. Haarnoja is a co-author of the original Soft Actor-Critic research paper. Their backgrounds help explain why robot learning, rather than a single fixed application, sits at the center of Eka's company-building effort.

Reinforcement learning gives a system a way to improve through interaction and feedback about outcomes. Training in a simulated environment lets a developer explore physical situations before asking hardware to encounter them. That changes a resource decision inside the company: how much learning to obtain from a person performing a task, and how much to generate through computation.

A simulator also makes a promise that the physical world can decline to honor. Contact depends on material properties, friction and the behavior of the machine itself. An automation buyer therefore needs to understand what carries over from training into its proposed task. Strong research credentials explain the team's approach; they do not substitute for defining performance at the installation.

Eka reports laboratory picking at 1,500 operations per hour. The conditions belong beside that figure: it is a company-reported laboratory result, rather than a general production guarantee. The more useful commercial discussion concerns how performance holds together as the object mix, task and operating conditions change.

A broader robotics market, one deployment at a time

DevCuration's coverage of Dream Labs' robot foundation-model API follows a related question about making robot capabilities available to other builders. The two companies should not be treated as interchangeable offerings. For Eka, the verified source record centers on force-aware manipulation and the work of carrying those skills into customer deployments.

The buyer's responsibility can become clearer as a supplier defines what it delivers. A manufacturer considering Eka would want to know how a task is specified, how changes are handled and what support accompanies the machine's learning. These are purchasing questions raised by the proposition, not commercial terms Eka has publicly promised.

The company's current careers destination keeps recruitment beside that deployment ambition. The Series A supports an organization trying to extend physical competence across more work, with research and customer requirements continuing to meet inside the same business. As Agrawal and Haarnoja expand the model's reach, the next installation brings another set of parts and working conditions for the team to carry into its learning.