The company, founded in 2023, secured $405m to develop general-purpose AI models designed to control mixed robot platforms, from humanoids to autonomous vehicles.
FieldAI has raised $405m across two recent rounds to develop foundation models for robot control. The latest $314m round closed in August. It was co-led by Bezos Expeditions, Prysm and Temasek. Axios reports a $2bn post-money valuation for the company.
Investors named across the two rounds include Khosla Ventures, Intel Capital, Canaan Partners, NVentures, BHP Ventures and Emerson Collective. FieldAI also lists prior backers Gates Frontier and Samsung. The company is based in Irvine, California, and was founded in 2023.
FieldAI builds “Field Foundation Models,” physics-based software for embodied AI. The models are designed to run across different robot types and tasks. Target platforms include humanoids, quadrupeds, mobile robots and autonomous vehicles.
The firm’s approach is hardware-agnostic. Its control unit can be installed on third-party robots that already support joystick operation, replacing manual teleoperation. The aim is to standardise autonomy across mixed fleets without redesigning mechanical platforms.
The product design adds a physics layer to the learning system. This gives the controller a second source of information beyond pattern recognition. The company frames this as a way to measure confidence and align actions with a user-defined risk threshold. The approach is intended to reduce unsafe actions by the agent during unseen tasks.
The company’s contracts span across construction, energy and urban delivery. The company positions its physics layer as a way to manage risk and uncertainty in changing environments. The design seeks to provide confidence estimates for actions, with thresholds set to match operational risk limits. The firm’s commercial focus is broad deployment of a single control stack across heterogeneous robot fleets.
