Generalist AI reached a valuation of US$ 3 billion after raising nearly US$ 200 million in a new funding round, according to people familiar with the transaction heard by TechCrunch. The round rapidly expands the volume of capital directed to the startup, which develops artificial intelligence models capable of controlling different types of robots.
The amount raised is backed by a document presented by Generalist itself to the Securities and Exchange Commission (SEC) of the United States. The Form D, filed on August 24, shows an offering of up to US$ 208.2 million, of which US$ 198.2 million had already been sold to 32 investors. The first sale occurred on August 7.
The SEC, however, does not disclose the company's valuation nor identify the round's investors. According to TechCrunch, citing two people with knowledge of the deal, the financing was led by 8VC and raised Generalist's value to US$ 3 billion. Axios had revealed a day earlier the raising of approximately US$ 200 million and also pointed to 8VC as the lead, although at that time it only reported that the valuation had surpassed US$ 2 billion.
US$ 600 million in a few months
The new investment comes less than three months after Generalist announced a round of US$ 400 million, led by Radical Ventures. At the time, the company said the investment had raised the total raised since its founding to more than half a billion dollars. Among the investors were 8VC, Union Square Ventures, Hanabi Capital and Norwest, in addition to earlier backers such as Nvidia, Spark Capital and Bezos Expeditions.
According to TechCrunch sources, that round valued Generalist at about US$ 2 billion. The new fundraising works as an extension of the same Series B, raising the combined size of the round to approximately US$ 600 million and increasing the company's valuation by about 50% in a few months.
The speed of the valuation increase highlights investors' growing interest in so-called physical AI, a field that seeks to bring the advances of large artificial intelligence models to machines capable of perceiving and acting in the real world.
The bet on a 'brain' for different robots
Founded in 2024 by Pete Florence and Andy Zeng, former researchers at Google DeepMind, and Andrew Barry, a former engineer at Boston Dynamics, Generalist follows a different strategy from companies focused on building humanoid robots. Its main product is the intelligence that controls the machines, and not necessarily the hardware.
The company describes its mission as building 'general intelligence for the physical world.' The idea is to develop models capable of operating different types of robots, environments and tools, reducing the need to create a specific system for each task.
This concept gained a new demonstration in August with the GEN-1.5, the company's latest model. According to Generalist, the system can learn certain physical tasks from a single demonstration with only 3 to 12 seconds in duration, without the need for a new traditional training process.
In tests released by the company itself, GEN-1.5 achieved an average success rate of 59% when receiving just one demonstration of different tasks. With five minutes of data and ten additional adaptation steps, the average rate reached 83%. The experiments involved actions such as opening containers, manipulating zippers and using tools.
Generalist stresses that the tasks are still relatively simple and short-lived. The point considered most relevant by the company is the model's emergent ability to learn new physical behaviors with a very small amount of data.
Race for a universal robotics platform
The volume of capital raised by Generalist shows how the race for the next generation of robotics is shifting from hardware to the models that control the machines.
Just as language models have come to perform very different tasks using a single base architecture, physical AI companies are trying to create systems capable of commanding different robots without each application having to be programmed virtually from scratch.
For investors, the possibility is to create a reusable intelligence layer in industrial arms, specialized machines and future general-purpose robots.
It is still early to know whether this approach will manage to reproduce in robotics the leap seen in generative models. Generalist itself acknowledges that its systems still operate on limited tasks. But the nearly US$ 200 million raised just weeks after a US$ 400 million round show that part of the venture capital market is willing to finance this bet on a large scale.
With a reported valuation of US$ 3 billion just two years after its founding, Generalist quickly joins the group of startups vying for the creation of the intelligence that could operate the next generation of robots.
