Generalist AI Pulled In Another $400M — Nvidia, Bezos and Fei-Fei Li Are Betting on 'AGI With a Body'

Big money piled into 'AGI with a body' again — this time $400M

Here's the deal: on June 4, robotics foundation-model startup Generalist AI raised $400 million in new funding at a $2 billion post-money valuation. Radical Ventures led the round, with 8VC, Union Square Ventures, Norwest, and Hanabi Capital joining. The returning-backer list is a flex — NVIDIA and Bezos Expeditions bet again, and new angels include AI researcher Fei-Fei Li, Xiaomi co-founder Bin Lin, and Naval Ravikant.

What the company builds is "physical AGI" — not smarts trapped in a chat window, but general intelligence that lets a robot handle complex tasks in the real physical world. The round pushes total funding past $500 million. The money goes into robot-learning models, physical data-collection infrastructure, compute, and expanding commercial deployments.

Why does it matter? Because it's a clear signal that the AI investment theme of 2025-26 is shifting from "software" to "the body." After chatbots and coding AI swept through, capital is now flowing toward giving robots hands and feet. Generalist AI's $400M is one of the sharpest pieces of evidence for that turn.

The cast — who's who

Start with Generalist AI itself, a startup building foundation models for robotics. The core idea: just as GPT got smart on text, a robot can get generally smart by training on vast "physical data." Not a robot that does one task well, but a "generalist" robot that adapts to new tasks on its own — hence the name.

Radical Ventures, which led the round, is an AI-specialist VC. Toronto-based, tightly linked to AI godfather Geoffrey Hinton, it bet early on foundation models and robotics. Its leading this round signals that the investors who understand AI best see "physical" as what's next.

Returning backers NVIDIA and Bezos Expeditions matter too. NVIDIA owns both the robot's brain (GPUs) and the simulation platform (the Isaac line), so it profits most when physical AI takes off — which is why it salts a stake into every promising robotics startup. Bezos has aggressively backed frontier tech through his personal vehicle, with robotics a core interest.

Finally, the new angels are symbolic. Fei-Fei Li opened the deep-learning revolution with ImageNet and now champions "spatial intelligence" — AI that understands the physical world. Her joining as an angel is effectively an academic stamp of credibility. Add Xiaomi co-founder Bin Lin and Silicon Valley legend Naval Ravikant and the list gets heavy.

What exactly happened

Let's get the facts straight. On June 4, Generalist AI announced a $400 million new round at a $2 billion post-money valuation (including the raise). Radical Ventures led; 8VC, USV, Norwest, and Hanabi joined new; NVIDIA and Bezos Expeditions returned. Total funding now exceeds $500 million. The plan: grow the robot-learning models, build physical data-collection infrastructure, secure compute, and scale commercial deployments.

The key phrase is "physical data." Text AI trains on the open internet; robots have no such corpus. The bodily experience of picking up an object, opening a door, or righting a tipped cup isn't sitting on the web. So the real fight among physical-AI companies is "who collects the most, most-diverse physical data." That Generalist is spending a big chunk on data-collection infrastructure is telling — that's the actual moat in this field.

Item	Detail
Announced	June 4, 2026
Raise	$400M (new round)
Valuation	$2B post-money (incl. raise)
Total funding	$500M+
Lead	Radical Ventures
New backers	8VC, Union Square Ventures, Norwest, Hanabi Capital
Returning	NVIDIA, Bezos Expeditions
New angels	Fei-Fei Li, Bin Lin (Xiaomi co-founder), Naval Ravikant
Goal	Physical AGI — general robot intelligence for the real world
Use of funds	Robot-learning models, physical-data infra, compute, deployment

The picture: a proven AI investor (Radical) leads, an infrastructure heavyweight (NVIDIA) backs it, the avatar of frontier bets (Bezos) supports it, and academic authority (Fei-Fei Li) stamps it. Money, tech, and credibility lined up in one round.

What each side gets

Generalist AI is the most direct beneficiary. $400M lets it push hard for a while on physical data collection and compute — the two most expensive parts of physical AI. And names like Fei-Fei Li and Naval help recruiting; top robotics researchers pay attention just because "Fei-Fei Li is involved." That halo of credibility is worth more than the cash.

NVIDIA is again the "pickaxe seller." Like the merchants selling shovels and jeans in a gold rush, NVIDIA profits no matter which physical-AI company wins, because those robots train on its chips and simulation platforms. Salting a stake into each promising startup is a strategy of buying its future customers early — a bet with little downside.

Across investors, this round is a bet on the "physical-AI thesis." Text and image AI are already dominated by Big Tech with sky-high valuations, leaving little room for new entrants. Physical AI is still early, so the chance to catch "the next OpenAI" remains. That's why specialist VCs like Radical lead and the USVs and Norwests line up behind.

History check — the humanoid boom, wins and losses

Big money into physical AI isn't new. On the success side are this category's giants. Humanoid startup Figure raised mega-rounds at multibillion valuations, and the broader NVIDIA-backed physical-AI ecosystem grew fast. In that wave, "robot foundation models" became a real category, and Generalist AI rode to the top of it. The narrative — port the "foundation-model scaling" logic that worked on text over to robots — sold well to investors.

But robotics history is full of glamorous failures. The painful one is Rethink Robotics. Its "Baxter" collaborative robot was once an icon of the robot revolution, but when the tech couldn't meet real-world expectations, it shut down in 2018. Google once hoovered up many robotics firms (Boston Dynamics included) and then unwound most of them — a similar lesson. Robots demo beautifully, but the road to a product that actually makes money is long and brutal.

The core lesson: physical AI has the biggest gap between demo and reality of any field. A stage video of a robot pouring coffee is slick; deploying it across thousands of factories and homes reliably is a completely different problem. With money but without crossing that gap, you vanish like Rethink. So Generalist AI's real test is "can it burn $400M to get past demos all the way to commercial deployment?" A flashy cap table and a successful product are separate things.

Competitor counter-plays — how do the others respond?

First, direct rivals. Generalist isn't the only one chasing robot foundation models. Startups like Physical Intelligence and Skild AI carry similar visions and raise huge rounds. Their counter-play is "data advantage" and "domain land-grab." Whoever first collects enough physical data to ship a genuinely usable robot policy wins. If Generalist scales data infrastructure, rivals are forced to pour money into the same place.

Big Tech's in-house robotics teams are a variable too. NVIDIA pushes humanoid platforms like Isaac and GR00T, Tesla builds Optimus, and Google DeepMind continues robotics research. These giants have overwhelming capital, compute, and talent, so startups can't fight them head-on. That's why a company like Generalist must differentiate on a niche Big Tech won't do directly, or on the generality of the model itself. NVIDIA investing in Generalist reflects its own bet that backing promising startups beats doing everything in-house.

Chinese robotics can't be ignored either. Companies like Unitree are rapidly deploying humanoid hardware on price. As hardware gets cheap, demand for the "intelligence" to sit on top explodes — exactly Generalist's target market. So China's hardware surge is both threat and opportunity for Generalist: it can sell brains to robots whose bodies just got cheaper.

So what actually changes — by audience

For robotics and AI founders, the key signal is "money is really flowing into physical AI." Investors wary of stretched text-AI valuations are eyeing physical as the next frontier. Whether you're in robot data, simulation, or hardware, a position on this wave means a friendly fundraising environment. But if your demos shine while your commercialization path is fuzzy, later rounds will filter you out.

For investors, weigh both sides: "physical AI is early so the upside is big, but the risk is too." Capital cycles are longer than text AI, and the hardware and real-world variables make failure rates high — Rethink Robotics is the proof. Even with a flashy cap table, stay careful until "actual deployment metrics" appear. A $2B valuation is a price on a promise, not on revenue.

For general observers, grab the big picture: "AI is starting to get a body." For years AI lived in the digital world of words, images, and code; now there's a serious push to load that intelligence onto robots and send it into the physical world. Fei-Fei Li's emphasis on "spatial intelligence" is the same thread. If it works, it reshapes factories, logistics, and even chores; if it fails, another "robot winter" could come. Generalist AI's $400M is one scene in that vast experiment.

FAQ

Q. What exactly is "physical AGI"? Different from just a robot? A. Unlike conventional robots that do one task well, it means general robot intelligence that adapts to new tasks on its own. Just as GPT is general on text, the goal is intelligence that's general in the physical world. Remember it's an aspiration, not a finished product.

Q. A $2B valuation means it makes that much revenue? A. No. $2B is the "price of a promise" investors put on future potential. Physical AI is early-stage, so most firms have no meaningful revenue yet. Don't confuse valuation with actual business results.

Q. Why does NVIDIA invest in all of these? A. Whichever physical-AI company wins, those robots train on NVIDIA chips and simulation platforms. It's the "pickaxe seller" strategy — salting stakes into promising startups means NVIDIA profits as the ecosystem grows, a bet with little downside.

Q. Is success guaranteed? A. Not at all. Robotics has the biggest demo-to-reality gap of any field, and plenty of well-funded firms (like Rethink Robotics) still failed. A flashy cap table is a credibility signal, not a guarantee of commercialization. The crux is "can it get past demos to real deployment?"

References

• Nvidia-Backed Robotics Startup Generalist AI Valued at $2 Billion — Bloomberg
• Generalist AI raises $400M at $2B valuation to build general intelligence for robotics — SiliconANGLE
• Generalist AI Raises $400M in New Funding to Develop Physical AGI — The AI Insider
• Generalist AI raises $400M to accelerate robot foundation models — Robotics & Automation News
• Generalist AI Raises $400M, Reaches $2B Valuation — Ventureburn

This article is not investment advice. The raise, valuation, and investor list reflect company announcements and press reports and may differ from reality; valuation in particular is a private-market estimate and highly volatile. Make decisions at your own discretion and risk.