Aiming at physical AGI, Mind Robotics, founded by the founder of Rivan, has raised $500 million.

In the past two years, during the investment boom in embodied intelligence, the mainstream narrative has centered around humanoid robots. Areas such as models, bodies, dexterous hands, and perception have all become hot topics.

Industrial robots seem to have become the technology of the previous generation and are part of the traditional products to be disrupted.

However, in fact, beyond humanoid robots, industrial robots still have significant room for improvement, and their scene requirements and commercialization potential are quite clear.

Many of the founders newly entering the industrial robot field have experience in the manufacturing industry. For example, RJ Scaringe, the founder of the overseas emerging electric vehicle manufacturer Rivian, founded a robot company called Mind Robotics in 2025. This company focuses on the industrial field and aims to meet RJ Scaringe's high comprehensive requirements for intelligence, precision, and reliability.

It recently received a $500 million Series A financing co-led by Accel and Andreessen Horowitz. In late 2025, it received a $115 million seed round financing led by Eclipse Capital. Currently, its valuation has reached approximately $2 billion.

If there are no robots on the market that meet the requirements, then build your own

Mind Robotics is RJ Scaringe's third company. He founded the electric vehicle company Rivian in 2009, and the company went public in 2021. In the first half of 2025, he spun off a company called Also, which develops small autonomous delivery electric vehicles and cooperates with the food delivery company DoorDash.

RJ Scaringe, Image source: Rivian

In the second half of 2025, he founded Mind Robotics. This company is not a horizontal expansion of a business but has emerged from identifying pain points and needs within his business.

Rivian's first released models were the high - priced R1 series. They have delivered more than 100,000 vehicles in total, which can be regarded as solving the problem of whether they can build cars. Similar to Tesla, it has passed the Model S stage.

Its second model, the R2, is a mid - size SUV. It is a high - volume model and has been pre - ordered tens of thousands of times. It will be delivered on a large scale in the first half of 2026.

Rivian encountered a problem that was exactly the same as what Tesla faced when building the Model 3: the "hell" of production capacity ramp - up.

Back then, Musk often slept in the factory to solve this problem; RJ Scaringe's idea was to use robots to manufacture cars.

So he began to observe industrial robot products on the market. He found that "traditional" industrial robots excel at handling repetitive and dimensionally stable tasks in strictly controlled environments; however, they cannot perform most of the truly complex tasks in the factory environment. Because these tasks require dexterous operations, the ability to adapt to changes, and physical reasoning abilities in a dynamic environment.

For example, the shape of wire harnesses is different every time; materials can bend; and components can shift. In the factory, these so - called edge cases are actually the norm.

Most of the emerging embodied intelligence companies focus on home or personal scenarios. Their robots are researching how to fold towels, wash dishes, and provide emotional value to people; only a small number of them focus on the industrial field.

Therefore, to solve Rivian's actual problems and unable to find a suitable product on the market, RJ Scaringe chose to found Mind Robotics.

Currently, there is no clear information about the specific products of Mind Robotics. However, based on the information revealed by RJ Scaringe, it is very likely to be a more intelligent robotic hand.

RJ Scaringe believes that hands are crucial in industrial manufacturing. In industrial manufacturing, for a robot to perform tasks similar to those of a human, in essence, it needs to have hands. The end of the hand needs to have a perception model and the ability to adjust on the X, Y, and Z axes. From the perspective of the robot system, most of the other technological efforts are to send the hand to the correct position (including mobility).

When considering the large - scale deployment of robots in the factory environment, it should minimize complexity, reduce failure modes, and lower power consumption. So humanoid robots are not necessary in this environment.

Because in the factory environment, the required operational design domain (ODD) is completely different from that in the home environment. Robots do not need to deal with stairs, there is no transition from carpet to tile flooring, there is no need to worry about being tripped by a cat, or stepping on a child. Designers can also map the environment because once deployed, the overall environment changes little.

To build a robot capable of handling dexterous, variable, and reasoning - intensive tasks, a complete robot platform must be constructed: an integrated model trained with real - world data, hardware designed for fine - control and robustness, and deployment infrastructure that supports continuous learning and iteration. This requires designing intelligence, machines, and the industrialization process as a whole.

Building such a platform cannot be accomplished by companies that have never industrialized products, have never really spent time understanding industrial operations, do not have a data flywheel for training models, and do not have a mature supply chain. RJ Scaringe founded Mind Robotics based on his experience at Rivian. He can cooperate closely with Rivian and obtain rich and diverse production data from its actual production line.

This injects power into the robot data flywheel and enables rapid iteration; and the improvement of the model can be directly deployed in actual operations. It is reported that Mind Robotics will deploy a large number of robots by the end of this year. In addition to Rivian, it should also cooperate with other manufacturing companies with needs.

Why is embodied intelligence in the industrial field an effective path to physical AGI?

There is more than one way to explore AGI. In - depth exploration in suitable niche fields may also lead to AGI. For example, in the field of large models, coding has been regarded as a popular route to explore AGI in the past two years. Because coding tasks have characteristics such as high complexity, verifiable feedback, and the ability to iterate repeatedly, they can form an iterative flywheel that uses intelligence to enhance intelligence.

Coincidentally, in the field of physical AI, the industrial environment also has similar characteristics. Continuous progress in intelligence in the industrial manufacturing field may lead to general physical artificial intelligence one day.

Breaking it down, deploying robots in the industrial environment generates data. The data improves the model and control system, and the performance improvement unlocks more extensive deployment, which in turn generates more data.

Companies that can closely integrate hardware, software, and manufacturing into a unified system are the ones that can build the most barriers and create hard - to - replicate competitive advantages.

Therefore, companies that already have such a foundation have started to move towards embodied intelligence. Tesla was the first to practice this. In 2021, it announced that it would develop robots and apply them to industrial production.

Chinese electric vehicle manufacturers such as XPeng and Li Auto have also deployed in the field of embodied intelligence.

The underlying idea of these manufacturers' deployment of industrial robots and embodied intelligence is that they have scenarios, data, and manufacturing capabilities, which can form an iterative flywheel.

Further, the current development direction of the entire AI industry has shifted from pure cognition to execution, whether it is the agents in the digital world or the embodied intelligence in the physical world.

The breakthrough point of embodied intelligence is not only to make the model understand the real three - dimensional world but also to manufacture machines that can act reliably in the physical world. To achieve this goal, data is definitely the foundation of everything, but this is a problem of the entire system's collaboration and requires the joint efforts of the entire entrepreneurial and innovative ecosystem. We are looking forward to seeing innovation in this field break through.

This article is from the WeChat official account "Alpha Commune" (ID: alphastartups). The author is Alpha Commune, which discovers extraordinary entrepreneurs. It is published by 36Kr with authorization.