General Intuition Raises $6B to Teach AI About Physics

General Intuition, a startup promising to build foundation models that teach AI agents to navigate space and time, is raising capital at a $6 billion pre-money valuation. The round is being backed by Valor Ventures, Point72 Ventures, and Seven Seven Six—three names that have collectively written enough checks to fund a small nation's defense budget. At six billion dollars, General Intuition is now valued higher than Mobileye before its IPO, higher than Figma at its peak, and roughly equivalent to the GDP of Iceland. All for a company whose core product is, stripped of the foundation-model language, a very expensive physics simulator.

General Intuition's stated mission is to train generalized AI agents how to move through space and time. This is, in plain English, robotics software. But calling it a "foundation model that trains generalized AI agents" sounds far more defensible in a pitch meeting than admitting you're building the same motion-planning algorithms that Boston Dynamics, Tesla, and Figure AI have been working on for a decade. The startup has not disclosed revenue, customer wins, or deployed units. No public information suggests they have anything other than a compelling narrative and a whitepaper. Yet here we are: six billion dollars, no product announcements, no revenue figures, and three sophisticated investors apparently convinced that this particular approach to spatial reasoning is categorically different from every other approach currently being funded.

Point72 Ventures, the venture arm of Steve Cohen's hedge fund, has a documented appetite for moonshots in AI and robotics—they backed Archer Aviation, which is still trying to figure out how to build electric aircraft that work. Valor Ventures has participated in some of the industry's most confident bets on foundation models and agents. Neither firm is known for sloppy due diligence, which makes this valuation either visionary or a cautionary tale waiting to be written. The fact that both firms are comfortable with this number suggests either that General Intuition has shown them something genuinely novel, or that the foundation-model-for-everything thesis has calcified into religious conviction.

The press materials almost certainly describe this as a "breakthrough in embodied AI" and a "paradigm shift in agent reasoning across spatial domains." Translation: we built a simulator that handles physics better than existing open-source options, and we're betting that's worth six billion dollars. The "foundation model" framing is crucial here—it allows the startup to borrow credibility from the GPT/LLM playbook, where early bets on general-purpose models did eventually create trillion-dollar value. By contrast, robotics is a domain where general-purpose approaches have historically lost to specialized ones, and where customer acquisition has been brutally slow.

The risks are straightforward: robotics deployment timelines are measured in years, not quarters; the competition includes well-funded specialists and massive incumbents; and the gap between "impressive physics simulation" and "useful robotic agent" has proven to be the hardest mile in the entire field. If General Intuition hits a typical robotics company's 18-36 month development cycle, they will need to deploy this technology at massive scale just to justify the valuation burn rate. Investors in foundation models have learned to live with losses, but robotics investors still expect hardware to move and customers to pay.

This deal is a perfect encapsulation of 2026 VC logic: take a known problem (robot motion planning), wrap it in trendy language (foundation models, generalized agents, spatial-temporal reasoning), add three credible investors, and price it like the next OpenAI. The only variable left unknown is whether the company can actually build something that works. In venture capital, that variable has never stopped a six-billion-dollar valuation before.

Six billion dollars to teach artificial intelligence something humans learned through evolution. What could possibly go wrong?