Generalist Robotics Hits $3B Valuation After $200M Extension Round
Generalist, the AI robotics startup founded by ex-Google DeepMind researchers, raised ~$200M led by 8VC, lifting its valuation to $3B after a $400M Series B.
Generalist, a robotics AI startup founded in 2024, has reached a $3 billion valuation after raising nearly $200 million in a funding extension led by 8VC, according to two sources familiar with the deal and a regulatory filing. The new capital is an extension of the $400 million Series B that Radical Ventures led in June at a $2 billion valuation, bringing the total round size to $600 million. The company is building a foundation model designed to control a wide range of robots, and recently released its Gen 1.5 model.
What happened
| Detail | Fact |
|---|---|
| New capital raised | Nearly $200M (extension round) |
| Lead investor (extension) | 8VC |
| Series B total | $600M |
| Original Series B | $400M led by Radical Ventures, June 2026 |
| Valuation at Series B | $2B |
| Current valuation | $3B |
| Founded | 2024 |
Generalist is building what the company calls an AI foundation model (a single large model trained to handle many tasks) that works across different robot hardware. Its latest release, Gen 1.5, lets robots learn new physical tasks from video demonstrations between 3 and 12 seconds long, according to the company’s own claims.
The startup was co-founded by Pete Florence and Andy Zeng, both former Google DeepMind researchers, alongside Andrew Barry, a former Boston Dynamics engineer. Early backers include Nvidia, Union Square Ventures, Bezos Expeditions, and AI researcher Fei-Fei Li, in addition to lead investors 8VC and Radical Ventures.
Generalist declined to comment. 8VC also did not respond to a request for comment, according to TechCrunch.
How does Generalist compare to its rivals?
| Company | Valuation | Backer(s) |
|---|---|---|
| Skild AI | $14B | SoftBank |
| Physical Intelligence | $11B (reported) | Not specified in source |
| Genesis AI | $3B (in talks, as of last month) | Not specified in source |
| Generalist | $3B | 8VC, Radical Ventures, Nvidia, others |
Generalist is working with a small number of customers and using their feedback to adapt the model for specific use cases, according to one source cited by TechCrunch. The startup operated largely under the radar until recently.
Why it matters
Investors are placing large bets on the idea that robotics is approaching its own “ChatGPT moment,” meaning a point where robots can handle general tasks without being programmed for each one individually. The logic is appealing: if software models can generalise across tasks, maybe physical machines can too.
The catch, as some VCs openly acknowledge, is that robots cannot be trained on the vast corpus of internet data the way large language models can. Physical world data is expensive, slow to collect, and far more limited in volume. That gap means a truly general robotics model could still be years away, even as valuations surge now.
For businesses watching this space, particularly those in warehousing, manufacturing, or field services, the Gen 1.5 claim of learning from short video clips is the most practically relevant detail. If it holds up at scale, the cost of deploying robots to new tasks drops significantly. But “claims” and “at scale in production” are very different things.
The rapid rise in valuations across this sector mirrors the early enterprise AI integration wave, where capital moved ahead of proven deployments. Businesses considering robotics automation should track which of these foundation model vendors can show real customer results, not just investor confidence.
Our take
The valuation jump from $2B to $3B in roughly two months is striking, but the numbers around it tell a more grounded story. A $600M round for a company founded in 2024 with a handful of customers reflects investor positioning for a market that does not yet fully exist, not current revenue. That is not necessarily a problem, but it is worth naming clearly.
The 3-to-12 second video demonstration claim for Gen 1.5 is the most interesting technical detail here. If verified independently, it would represent a real reduction in the cost of deploying new robot behaviors. Right now, though, it is a company claim with no third-party benchmark attached.
For our clients, the practical relevance of general robotics foundation models is still a few years out for most use cases. The nearer-term opportunity is in AI integration at the software layer, where similar “learn from examples” logic is already available and deployable today. The robotics layer will catch up, but we would not restructure operations around it yet.
If you want to stay current on how AI funding rounds translate into real tools and services, our AI news coverage tracks the practical signal through the funding noise.
Frequently asked questions
What is Generalist robotics and what does it do?
Generalist is a robotics AI startup founded in 2024 by former Google DeepMind researchers Pete Florence and Andy Zeng, along with ex-Boston Dynamics engineer Andrew Barry. It is building an AI foundation model designed to work across many different robot types, with its Gen 1.5 model enabling robots to learn tasks from video demonstrations as short as 3 to 12 seconds.
How much has Generalist raised and who are its investors?
Generalist has raised a total of $600M across its Series B, including a $400M raise led by Radical Ventures in June 2026 and a nearly $200M extension led by 8VC. Other backers include Nvidia, Union Square Ventures, Bezos Expeditions, and AI researcher Fei-Fei Li.
How does Generalist's valuation compare to other robotics AI startups?
At $3B, Generalist is smaller than Skild AI ($14B, backed by SoftBank) and Physical Intelligence (reported $11B). Genesis AI was in talks to raise at a $3B valuation as of last month, putting it at a similar level to Generalist.
What is a robotics AI foundation model?
A foundation model in robotics is a single large AI model trained to control many different types of robots across many different tasks, rather than a model trained specifically for one robot or one job. The goal is for the same model to generalise across hardware and tasks, similar to how large language models handle diverse text tasks.
