Whether you're looking for an angel investor, a growth advisor, or just want to connect — I'm always open to great ideas.
Get in TouchPhysical Intelligence, the two-year-old startup founded by ex-Google DeepMind researchers, is in talks to raise $1 billion at an $11 billion valuation — doubling its worth in just four months and signaling that physical AI has reached its inflection point.
AI, startups & growth insights. No spam.
TL;DR: Physical Intelligence (π), the San Francisco robotics startup founded by former Google DeepMind researchers, is in talks to raise $1 billion in new funding at a valuation exceeding $11 billion — effectively doubling its $5.6 billion valuation from just four months ago. Founders Fund is set to lead, with Lightspeed Venture Partners in discussions to join. If closed, the company will have raised over $2 billion since its March 2024 founding, making it one of the fastest-funded robotics startups in history.
A two-year-old company. A $1B fundraise. A valuation that doubled in four months. Physical Intelligence is not building a robot — it is building the AI brain that makes every robot smarter, and investors are now treating that distinction as worth $11 billion. Co-founder Sergey Levine describes the company's mission simply: "Think of it like ChatGPT, but for robots." In a field where Tesla is racing to deploy Optimus at scale and OpenAI is circling the physical world, Physical Intelligence just signaled that the software layer of robotics may be worth more than the hardware.
In November 2025, Physical Intelligence closed a $600 million funding round that valued the company at $5.6 billion, according to Bloomberg reporting at the time. That round was itself remarkable — a Series B for a company that had existed for only eighteen months, led by Alphabet's independent growth fund CapitalG, with participation from Lux Capital, Thrive Capital, and Amazon founder Jeff Bezos.
Four months later, the same company is now in discussions to raise another $1 billion at more than double that valuation. The pace is striking even by the standards of 2025-2026 AI funding, where valuations have expanded dramatically across the board. Several factors explain the acceleration.
Product velocity. Between November 2025 and March 2026, Physical Intelligence shipped multiple iterations of π0. The π0.5 update delivered better open-world generalization; π0.6 added reinforcement learning, allowing the model to improve autonomously from real-world experience. These are architectural advances, not incremental patches.
Market timing. The broader robotics investment environment re-priced sharply in early 2026. Tesla's Optimus production announcements, Figure AI's White House demonstrations, and a cascade of Chinese humanoid robot announcements compressed the perceived timeline for physical AI from "eventually" to "now."
Competitive positioning. Physical Intelligence is not competing to build the best robot body — it is the AI layer that sits on top of any robot body. That model-as-a-service approach is more defensible and more scalable than hardware.
The result: $5.4 billion in paper value creation in four months, without a single hardware product shipping.
Physical Intelligence was founded in March 2024 by a team that reads like a who's who of academic and industry robotics research.
Karol Hausman, CEO and Co-founder — Former Staff Research Scientist at Google DeepMind and adjunct faculty at Stanford University. Hausman specialized in manipulation learning — teaching robots to interact with physical objects — which is precisely the hardest problem in general-purpose robotics.
Sergey Levine, Chief Scientist and Co-founder — Professor at UC Berkeley whose lab essentially pioneered deep reinforcement learning for robotic manipulation. Levine's academic publication record includes foundational work on learning from demonstrations and meta-learning that underpins the π0 architecture. His framing of Physical Intelligence's goal as "ChatGPT, but for robots" is not a marketing line — it is a precise technical description of what the company is building.
Chelsea Finn, Research Lead and Co-founder — Associate professor at Stanford University, known for work on meta-learning and sim-to-real transfer. Finn's research on how models can quickly adapt to new tasks with minimal examples directly informs Physical Intelligence's approach to generalist policies.
Brian Ichter, VP Engineering — Former Google Research robotics veteran focused on optimal control and large-scale robotic experimentation. Ichter bridges the gap between research excellence and production-scale systems — a critical role for a company that needs both.
Lachy Groom, COO and Co-founder — Former Stripe product leader turned angel investor, responsible for fundraising, partnerships, and go-to-market. His presence is a deliberate signal that this company intends to build enterprise-grade infrastructure, not just publish research.
The founding team's composition — two Berkeley/Stanford professors, two Google DeepMind veterans, and a Stripe-trained operator — is unusually complete for an early-stage startup. Most academic spinouts lack business-building muscle; most industry spinouts lack foundational research depth. Physical Intelligence has both.
Released in February 2025 and continuously updated since, π0 (pi-zero) is Physical Intelligence's first generalist robot policy. Understanding what it actually does is essential to understanding the company's valuation.
Traditional robot software is task-specific: a robot arm programmed to pick up a red ball will fail if asked to pick up a blue cube. Physical Intelligence's approach is different — train a single large model on a massive, diverse dataset of robot behavior, then fine-tune it to new tasks with minimal additional data.
The technical details: π0 is a three-billion-parameter transformer model built on top of PaliGemma (Google's vision-language model), extended with robot state and action generation modules. It was trained on over 10,000 hours of real-world robot data spanning seven different robot embodiments and 68 distinct tasks. It uses flow matching to generate smooth, continuous action trajectories at high frequency — enabling dexterous manipulation that earlier approaches could not achieve.
What can π0 actually do? The demo tasks are instructive: folding laundry, cleaning tables, scooping coffee beans, peeling vegetables, assembling items. These are not warehouse automation tasks designed around a robot's strengths — they are household tasks designed around a human's world. The model handles unstructured environments, variable object placement, and material properties that rigid traditional software cannot accommodate.
The π0.6 update — adding reinforcement learning — is the most significant architectural development. It allows the model to improve through trial and error in real environments, not just pre-collected demonstrations. A robot policy that improves autonomously from experience is qualitatively different from one frozen at training time.
Physical Intelligence has also open-sourced the base π0 model via their GitHub repository, which has generated significant developer community activity. Open-sourcing the foundation while selling fine-tuning, deployment infrastructure, and enterprise support is a familiar playbook — it accelerates ecosystem development without giving away the commercial moat.
Physical Intelligence's capital trajectory is among the most aggressive in recent AI startup history.
| Round | Amount | Valuation | Date | Lead Investor |
|---|---|---|---|---|
| Seed | $70M | Undisclosed | 2024 | Lux Capital, Thrive Capital |
| Series A | $400M | $2.4B | Mid-2024 | Lux Capital, Thrive Capital |
| Series B | $600M | $5.6B | November 2025 | CapitalG (Alphabet) + Jeff Bezos |
| Series C (in talks) | $1B | $11B+ | March 2026 | Founders Fund + Lightspeed |
If the current round closes at the reported terms, Physical Intelligence will have raised more than $2 billion in under two years — faster than almost any robotics company in history, and at valuations that were previously reserved for companies with real revenue.
The investor roster rewards examination. Lux Capital and Thrive Capital participated in both seed and Series A, providing continuity. CapitalG — Alphabet's independent growth fund — led the Series B, signaling that Google believes Physical Intelligence's approach is either complementary to or competitive with DeepMind's own robotics work. Jeff Bezos' personal participation adds distribution optionality given Amazon's warehouse automation ambitions.
Founders Fund's reported participation brings Peter Thiel's firm into the cap table — known for backing deep technology companies that others consider too ambitious. Lightspeed Venture Partners, one of the most active enterprise infrastructure investors, joining alongside Founders Fund suggests this round is being positioned as a software infrastructure play, not a hardware bet.
According to Bloomberg's reporting and TechCrunch coverage, the new round is anchored by Founders Fund with Lightspeed Venture Partners in discussions to join, alongside returning investors Thrive Capital and Lux Capital.
Founders Fund's involvement is significant for reasons beyond capital. The firm was an early backer of SpaceX, Palantir, and Anduril — companies that bet on deep technology with long timelines and transformational ambitions. Their investment thesis has historically been that the best opportunities are in fields that mainstream investors consider too hard, too capital-intensive, or too long-dated. Physical Intelligence fits that profile precisely: general-purpose robot AI is technically extremely hard, requires sustained capital investment, and will take years to generate the revenue that justifies current valuations.
Lightspeed adds a different lens. The firm has one of the best track records in enterprise infrastructure — Snap, Affirm, Mulesoft, HashiCorp. Its interest signals that Physical Intelligence's models could become embedded in manufacturing and logistics contracts where switching costs are high and contract values are large.
For context: OpenAI's $40 billion round from SoftBank reset expectations for AI fundraising. Physical Intelligence at $1 billion is not an outlier — it is a data point in a new normal where credible AI infrastructure companies capitalize at levels previously reserved for pre-IPO megacaps.
Physical Intelligence is not operating in a quiet corner of the market. The physical AI space has become one of the most contested areas in technology, with well-funded competitors attacking from multiple directions.
Tesla Optimus — Elon Musk's humanoid robot program is the highest-profile project in the space. Tesla is planning to scale production to one million units per year by 2030, with a target price of $20,000-$30,000 per unit. Tesla's advantage is vertical integration: hardware, software, manufacturing, and distribution in one company. Its disadvantage is that its robot AI is designed for Tesla's specific hardware — it is not a general-purpose platform.
Figure AI — One of Physical Intelligence's most direct competitors in the US, Figure has raised significant capital and has demonstrated its robots at the White House. Figure's approach combines its own humanoid hardware with AI software, which creates both a competitive advantage (full-stack control) and a strategic limitation (hardware capital intensity).
Boston Dynamics — The Hyundai-owned pioneer has decades of hardware excellence but has historically struggled to translate that into scalable, updatable software — the precise gap Physical Intelligence is targeting.
Google DeepMind — Physical Intelligence's founders came from DeepMind, which continues to invest heavily in robotics research. The relationship is simultaneously origin story and competitive dynamic.
OpenAI — OpenAI has invested in Figure AI and is testing its multimodal models as robot brains. Its capital resources make it a credible long-term threat to any AI-first robotics company.
Chinese Competitors — Unitree, UBTECH, and dozens of smaller startups are driving hardware prices down aggressively — some humanoid robots now cost under $15,000. If they commoditize the hardware layer faster than expected, the software platform Physical Intelligence is building becomes even more valuable.
Morgan Stanley projects the humanoid robot market could reach $5 trillion by 2050. Even with heavy skepticism applied, the serviceable market for general-purpose robot AI in manufacturing, logistics, and healthcare is measured in hundreds of billions within a ten-year horizon.
The most important strategic question in physical AI is not which robot is the best — it is whether the AI software layer will be bundled with hardware or sold separately as a platform.
The historical analogy is smartphones. In 2007, every phone had its own custom OS. By 2012, iOS and Android had won, and hardware companies competed on how well they differentiated around the OS. The companies that tried to own both layers (Nokia, BlackBerry) largely lost to those that won on software (Apple) or provided the platform (Google).
Physical Intelligence is betting robot AI follows the same arc. Its π0 model runs on seven different robot embodiments — explicitly untied to any single hardware platform. The company is building what amounts to an operating system for robots: a general-purpose foundation that hardware makers can customize and update over time.
This positioning differs fundamentally from Tesla's, Figure's, and Boston Dynamics' approaches — all of which bundle hardware and software. If Physical Intelligence's bet is correct, those companies will eventually face pressure to license a platform or build their own, ceding the software layer.
The counterargument: tight hardware-software integration — the Apple model — may produce the strongest economic moat. If Tesla's Optimus is genuinely better because Musk controls both body and brain, platform providers face structural disadvantages.
At $11 billion on no meaningful revenue, Physical Intelligence's investors are betting the platform model prevails — the same bet every Android investor made in 2008.
An $11 billion valuation for a two-year-old company with no disclosed revenue demands scrutiny.
Revenue timeline. Physical Intelligence has disclosed neither revenue nor signed commercial contracts. The gap between research demonstrations and enterprise-scale deployment is significant — both technically and organizationally.
Hardware dependency. The π0 model runs on other companies' robots. If hardware partners build their own AI layers, or a competitor's model reaches equivalent performance, Physical Intelligence's value proposition weakens materially.
The RL race. Reinforcement learning for robotics is being pursued intensely by Google DeepMind, OpenAI, and multiple academic groups. Model capability leads in AI have proven short-lived, and Physical Intelligence needs to compound advantages faster than better-resourced competitors close the gap.
Regulation. Physical AI in healthcare, elder care, and food preparation will face regulatory scrutiny that software-only AI has largely avoided. Liability frameworks for robot errors do not yet exist, and their eventual form could constrain deployment timelines.
Valuation multiple. At $11 billion on no disclosed revenue, the multiple is essentially infinite — sustainable only if commercial inflection arrives soon and at scale. The funding rounds are growing faster than public evidence of commercial traction.
If the round closes, Physical Intelligence will have more capital than ever — and more pressure to prove the commercial opportunity matches the research ambition.
Near-term priorities will center on enterprise partnerships and commercial deployments. Physical Intelligence has been explicit that its target is industrial and commercial automation, not consumer robotics. Signed contracts with manufacturing, logistics, or healthcare companies would be the validation event investors are waiting for.
The talent build will accelerate. With approximately 80 employees and over $2 billion in total capital, the company is significantly under-staffed relative to its ambition. A large portion of the new round will go toward researchers and engineers who can maintain model development velocity.
An IPO or strategic acquisition is the implied endpoint. Alphabet, Amazon, and Microsoft are the most obvious strategic acquirers — each has strong incentives to own the AI layer of the robot economy rather than license it. An IPO becomes viable once commercial revenue substantiates the current valuation narrative.
The broader signal from this round is about timing. 2024 was speculation about physical AI. 2025 was demonstration. If Physical Intelligence's bets pay off, 2026 will be remembered as the year the physical AI era began in earnest — the moment the market stopped asking whether general-purpose robot AI was possible and started asking which companies would own it.
DeepMind alumni building $11 billion companies in two years is not a coincidence. It is the output of the world's most concentrated robotics research talent deciding that the timing, the technology, and the capital are finally aligned. Whether Physical Intelligence becomes the dominant robot AI platform — or Tesla, OpenAI, or an unnamed challenger gets there first — is the defining technology race of the decade.
What is Physical Intelligence? Physical Intelligence (π) is a San Francisco-based AI robotics startup founded in March 2024 by former Google DeepMind researchers and professors from Stanford and UC Berkeley. It builds general-purpose AI models that can power robots to perform a wide variety of physical tasks.
How much has Physical Intelligence raised in total? If the current round closes, Physical Intelligence will have raised more than $2 billion in under two years — covering a $70 million seed, $400 million Series A, $600 million Series B, and the reported $1 billion new round.
Who founded Physical Intelligence? Co-founders include Karol Hausman (CEO, ex-Google DeepMind), Sergey Levine (Chief Scientist, UC Berkeley), Chelsea Finn (Research Lead, Stanford), Brian Ichter (VP Engineering, ex-Google Research), and Lachy Groom (COO, ex-Stripe).
What is the π0 model? π0 is Physical Intelligence's generalist robot policy — a 3-billion-parameter transformer trained on over 10,000 hours of real-world robot data spanning seven robot embodiments and 68 tasks. The base model has been open-sourced on GitHub.
Who are Physical Intelligence's main competitors? Physical Intelligence competes with Tesla Optimus, Figure AI, Boston Dynamics, Google DeepMind, and OpenAI. Unlike most competitors, it focuses exclusively on the AI software layer rather than building its own robot hardware.
Does Physical Intelligence have revenue? Physical Intelligence has not publicly disclosed revenue or signed commercial contracts as of March 2026. The company remains primarily in research and development, with commercial deployments expected as a near-term priority given its enterprise-focused positioning.
Rhoda AI emerges from 18 months of stealth with a $450M Series A at $1.7B valuation — training robots on hundreds of millions of internet videos instead of teleoperation, backed by Khosla Ventures and Premji Invest.
Sunday Robotics raises $165M Series B at $1.15B valuation, planning to ship consumer humanoid robots for under $25,000 by Thanksgiving 2026 — the first unicorn focused exclusively on household robotics.
Mind Robotics — spun out of electric vehicle maker Rivian — raises $500M in a Series A co-led by Accel and a16z at roughly $2B valuation, applying EV manufacturing expertise to build industrial AI robots.