Enact Intelligence Raises $30M for Force-Aware Robotics

Manuel Yves Galliker is giving his robotics researchers responsibility for the machinery that supplies their training observations. At Zurich-based Enact Intelligence, the founder and CEO is developing the moving hardware and the intelligence together, seeking useful information in the forces a robot encounters as it handles the world.

Enact announced a $30M pre-seed on October 22, 2026. Notion Capital says it co-led the round with redalpine, with firstminute capital participating; the company also names Yann LeCun, Arthur Mensch, Andrew Lonsberry and Thomas Wolf as angel investors. The money supports a development effort in which better physical observations could become an advantage for the model learning from them.

The prospective customer is buying into that relationship as much as a proposed robot capability. Enact's force-and-touch approach aims to make manipulation more adaptable, with the hardware, data and model teams working on the same problem. For industrial operators, the commercial question is how that combined supplier can carry a learning system from development into varied daily work.

What Enact's $30M pre-seed will fund

Enact will use the financing to expand its team and accelerate hardware and R&D development, according to Tech.eu's original funding report. The pre-seed label sits beside an engineering scope that reaches from actuators to model architectures. Keeping those programs connected requires people who can understand what an observation means mechanically as well as what it contributes to learning.

Notion Capital's company profile identifies Galliker as founder and CEO and lists 2026 as Enact's founding year. His experience includes ETH Zurich, research at Caltech, work on 1X's NEO humanoid and work at Physical Intelligence. That background helps explain the breadth of the assignment he has taken on in Zurich, where the company is building around both hardware and AI.

The financing gives Enact room to recruit across those disciplines while developing the system. Individual investors' checks and the company's valuation have not been disclosed in the reviewed announcements, so the $30M should be read as the announced round rather than a basis for estimating ownership or lifetime capital. The named angels invested as individuals; their involvement establishes no commercial partnership with their employers.

Why force information belongs in the learning process

Enact is working on an information problem inside a physical machine. Cameras supply visual observations, while force feedback can help describe the contact that occurs during movement. In a simple illustrative handling task, knowing where an object is and knowing how it resists movement answer different questions. Enact wants information from the latter available to the system choosing its actions.

Redalpine describes Enact's proprietary actuators as designed for torque transparency, feeding force and touch information into a foundation model alongside vision and language. This is a hardware design commitment as well as a model-development thesis. Treating the actuator as a source of observations makes its physical characteristics relevant to what the learning system can know.

That arrangement also changes the team's freedom to respond when the available information is inadequate. Researchers can examine whether a difficulty belongs in the model, the data pipeline or the machine producing the observations. Changing those pieces together is an organizational challenge, but it gives Enact a broader set of engineering decisions to make within one company.

The company's launch materials describe more efficient learning and faster task execution as intended benefits. They provide no numerical comparison that would let an outside buyer measure those gains. A prospective deployment therefore starts with an engineering proposition and a set of tasks to evaluate, rather than a published performance specification that settles the purchase.

Zurich brings the hardware work into the hiring plan

Enact's official team and hiring page lists experience from 1X, Google DeepMind, Physical Intelligence, Helsing, Amazon Robotics and other organizations. The company is recruiting as it builds a system spanning models, data, actuators and sensors. Locating that work together gives hiring an unusually direct connection to the product's architecture.

A model researcher and a hardware engineer can encounter the same failure from different ends. One sees an inadequate observation or an action that did not generalize; the other can inspect the mechanism that generated it. The commercial promise of putting those people close to one another is a shorter route from a physical problem to a decision about what should change. That is an interpretation of Enact's development approach, rather than a reported result from its deployments.

Industrial customers would eventually need that coordination to survive beyond the lab. A proposed improvement has to make sense for the equipment, the task and the people responsible for keeping work moving. Changing a machine to produce better training observations may create useful capability, while also raising questions about integration and support that belong in the customer's evaluation.

Enact has not named production customers or disclosed revenue in the launch sources reviewed. Those omissions leave the early commercial relationship open: which tasks will give the team useful feedback, and which operators will see enough value to help establish how the system should work in practice?

Robotics capital is financing different supplier relationships

Enact's full-stack approach places a particular boundary around the company. It wants influence over the physical observations, their path into training and the model that turns information into action. That gives a prospective customer a combined engineering counterpart, with the opportunity and responsibility that come from keeping those choices together.

Other robotics businesses are drawing their boundaries elsewhere. DevCuration's coverage of Dream Labs' robot foundation model API examines a service intended for other robotics builders. The two approaches give operators different supplier relationships to assess. The relevant comparison is the work each supplier takes responsibility for, rather than an assumption that one financing establishes a universal architecture for robotics.

For Enact, the current capital supports the people and hardware development behind its chosen boundary. The useful buying questions follow that boundary too: what information the machine provides, how the model uses it, and how improvements can reach equipment doing actual work. Public technical results and named deployments would give those discussions more substance as development proceeds.

Galliker's team is now recruiting into a program whose hardware decisions can shape its future learning. The industrial conversations ahead could influence which physical interactions become useful data, giving the first operators involved a role in the experience Enact's machines carry into subsequent tasks.