Home Bots & BusinessDyna Robotics Raises $120 Million for Robotic Foundation Models

Dyna Robotics Raises $120 Million for Robotic Foundation Models

by Pieter Werner

Dyna Robotics has raised $120 million in a Series A funding round to advance development of its robotics foundation models, the company announced . The round was led by Robostrategy, CRV, and First Round Capital, with participation from Salesforce Ventures, NVentures, the Amazon Industrial Innovation Fund, Samsung Next, and LG Technology Ventures.

The California-based company said the funding will support expansion of its research and engineering teams and accelerate work on production-ready general-purpose robots powered by proprietary embodied AI foundation models.

Dyna launched in 2024 with a $23.5 million seed round and has since released its DYNA-1 model, which the company describes as a general-purpose robotics foundation model. According to Dyna, the system has achieved a reported success rate of more than 99 percent in continuous 24-hour operations and has been deployed in commercial settings including hotels, restaurants, laundromats, and gyms.

The company said its approach relies on a single-weight foundation model designed to perform a wide range of tasks across different environments. Dyna’s co-founders include Lindon Gao, formerly of Caper AI; Jason Ma, a former DeepMind research scientist; and York Yang, also a co-founder of Caper AI, which was acquired in 2021 for $350 million.

In a statement, CEO Lindon Gao said that the company’s models are designed to adapt to new environments without additional data, while Jason Ma emphasized the importance of building systems that combine generalization with high performance in complex tasks. York Yang highlighted what he described as the convergence of advances in AI, robotics hardware, and labor demand as drivers for the company’s ambitions. Dyna said it intends to use the data collected from real-world deployments to continue improving its models as it works toward developing robots capable of what it terms “physical artificial general intelligence.”

Misschien vind je deze berichten ook interessant