Former NVIDIA Researchers Raise $90 Million to Build a Shared AI Brain for Low-Cost Robots

A robotics startup founded by former NVIDIA researchers, Dream Labs, has raised $90 million to develop artificial intelligence technology designed to help robots operate more effectively across different tasks and environments.

The company is part of a growing wave of robotics businesses seeking to make intelligent machines more affordable and practical for commercial use. Rather than developing a separate control system for every robot and application, the startup is working towards technology that can give different machines a common foundation for understanding and performing physical tasks.

The funding highlights continued investor interest in robotics, particularly systems that combine artificial intelligence with the physical world.

Moving beyond specialised robots

Robots have been used in factories, warehouses and industrial facilities for decades. Many traditional systems are programmed to perform specific operations under carefully controlled conditions, such as moving components along an assembly line or transporting materials between fixed points.

Although these machines can be highly effective, changing their tasks or operating environments can require significant engineering work. A robot designed for one activity may struggle when presented with unfamiliar objects, different surroundings or instructions outside its original programming.

Artificial intelligence is creating new opportunities to address these limitations.

Modern AI systems can process visual information, interpret instructions and help machines respond to changing conditions. Researchers are increasingly exploring whether these capabilities can be combined into general-purpose systems that allow robots to learn a wider range of activities.

Dream Labs is pursuing this direction by developing a shared AI foundation intended to support multiple robotic applications.

Why a shared AI foundation matters

A shared intelligence system could potentially reduce the amount of specialised software required to deploy robots.

For example, a warehouse robot might need to identify packages, navigate aisles and move objects safely. A robot operating in a manufacturing facility could require different physical capabilities, but might still benefit from common systems for perception, planning and interpreting instructions.

If a common model can handle these underlying functions, developers may be able to adapt it to different machines rather than build every capability from scratch.

This could make robotics development more efficient and create opportunities for smaller companies that cannot afford to develop sophisticated AI systems independently.

However, a shared AI model does not automatically make every robot capable of performing every task. Machines have different sensors, motors, physical designs and operating limits. Software must still be adapted and tested to ensure that robots behave safely in their intended environments.

The challenge of bringing AI into the physical world

Training AI for robotics presents challenges that do not always arise in conventional software applications.

A chatbot can generate an incorrect response without necessarily causing a physical incident. A robot acting on an incorrect prediction could damage equipment, drop an object or endanger someone nearby.

Robotics systems must therefore combine intelligent decision-making with reliable sensing, physical control and safety mechanisms.

Developers also need sufficient training data covering the variety of objects, environments and situations a robot may encounter. Real-world testing can be expensive and time-consuming, while simulated environments may not fully reproduce the complexity of physical operations.

Dream Labs will need to demonstrate that its technology can work reliably outside controlled demonstrations and provide measurable benefits to customers.

Why investors are interested

The $90 million funding round reflects expectations that AI could expand the range of tasks robots can perform.

Businesses face pressure to improve productivity, address labour shortages in certain industries and make operations more flexible. Robotics could help with repetitive, physically demanding or hazardous activities, provided the technology is reliable and economically viable.

A shared intelligence platform could also create a different business opportunity from manufacturing robots directly. By supplying underlying software and models, a company could potentially work with multiple hardware manufacturers and serve several industries.

The commercial opportunity will depend on the quality of the technology, the cost of deployment and the availability of customers willing to adopt it.

What comes next for Dream Labs?

The new investment gives Dream Labs resources to advance its technology and compete in a field attracting substantial attention from technology companies and investors.

Its progress will depend on demonstrating practical performance, including how well its systems transfer between tasks, how much training is required and whether they can operate safely in real-world settings.

Competition is also intensifying as established robotics manufacturers, AI laboratories and new startups develop systems designed to make machines more adaptable.

If Dream Labs succeeds, its technology could help lower some of the barriers to deploying intelligent robots. But the industry still has significant technical and commercial challenges to overcome before general-purpose robotics becomes commonplace.