Physical Intelligence Raises $1B, Valuation Doubles to $11B in Four Months (2026)

Physical Intelligence, the San Francisco AI robotics startup that builds foundation models for physical robots, is in talks to raise approximately $1 billion in new funding at a post-money valuation exceeding $11 billion, according to a Bloomberg report from March 27, 2026. The deal would effectively double the company's $5.6 billion valuation achieved just four months prior — one of the fastest valuation increases ever seen in the robotics sector.

What Happened

Bloomberg reported on March 27, 2026, that Physical Intelligence (stylized as π, and operating at pi.website) is in advanced discussions with investors including Founders Fund, Thrive Capital, and Lux Capital for a new round of approximately $1 billion. The company has not officially confirmed the round. If closed at the reported terms, the deal would bring Physical Intelligence's total funding to well over $1.5 billion — on top of its prior $400 million Series B raise in November 2024 that set the original $5.6 billion valuation.

Physical Intelligence was founded in 2024 by a team that includes Chelsea Finn (Stanford AI Lab), Sergey Levine (UC Berkeley Robotics and AI Lab), and Karol Hausman (formerly of DeepMind). The company's core thesis is that robotics needs a "foundation model moment" — the same shift that large language models brought to text — to enable robots to generalize across tasks and physical environments.

Key Details

• Funding target: ~$1 billion — according to Bloomberg, the largest single robotics AI round in 2026 so far
• New valuation: >$11 billion — up from $5.6 billion in November 2024, a doubling in under four months
• Investors in discussions: Founders Fund, Thrive Capital, and Lux Capital, joining existing backers Bond, Jeff Bezos, Khosla Ventures, OpenAI, Redpoint Ventures, Sequoia Capital, and CapitalG
• Latest model milestone: In March 2026, the company published research on "Precise Manipulation with Efficient Online RL," using RL tokens to improve real-world robotic task throughput
• Open-source contribution: Physical Intelligence open-sourced the weights and code for their π0 model in February 2025, with the π0-FAST autoregressive variant — a rare move for a well-funded AI robotics company
• Revenue trajectory: Not publicly disclosed, but investor interest at this pace of valuation growth implies significant progress on commercialization

About Physical Intelligence and the π0 Model Family

At the core of Physical Intelligence's technology is the π0 (pi-zero) family of vision-language-action (VLA) models. π0 was first announced in October 2024 as a 3-billion parameter model that combines internet-scale vision-language pretraining with multi-robot physical training data. On benchmark household tasks, π0 achieved near-perfect scores on laundry folding and table-bussing — tasks where prior models (OpenVLA, Octo) scored zero.

π0.5, released in April 2025, extended the model to open-world generalization — enabling robots to clean up an entirely new kitchen or bedroom it had never been trained in, following language instructions. π0.6 (November 2025) added reinforcement learning from real-world experience, improving throughput and success rates on precise manipulation tasks. The latest research (March 2026) introduces multi-scale embodied memory for tasks longer than ten minutes.

What Developers and Researchers Are Saying

Reaction in the robotics and AI research community has been strongly positive. The rapid pace of progress — from a single generalist policy (October 2024) to open-world generalization (April 2025) to RL-trained models (November 2025) in just over a year — is being cited as evidence that the "foundation model approach" to robotics is working. The open-sourcing of π0 weights has generated significant interest in academic robotics labs. Critics note, however, that current demos still rely on expensive hardware, and the gap between controlled lab performance and robust real-world deployment remains a challenge for the entire field.

What This Means for Developers

For developers and researchers working in robotics and embodied AI, Physical Intelligence's rapid valuation growth signals where AI investment is heading in 2026: from language and images toward physical-world interaction. The open-sourced π0 and π0-FAST models are available at github.com/Physical-Intelligence/openpi for experimentation. The company's growing dataset and compute advantage may make it increasingly difficult for academic research labs to keep pace on the full-system benchmarks — though the open-sourcing commitment at least keeps the underlying architectures public.

What's Next

If the $1B round closes as reported, expect Physical Intelligence to accelerate both hardware partnerships and deployment with commercial robotics customers. The company has not disclosed an IPO timeline, but at an $11 billion valuation with this fundraising trajectory, public market discussions are likely within the next 18-24 months. Watch their blog at pi.website/blog for new model releases and technical papers.

Sources

• TechCrunch: Physical Intelligence reportedly raising $1 billion again — Primary news coverage citing Bloomberg reporting
• Physical Intelligence Official Blog — Model release announcements (π0, π0.5, π0.6, online RL research)
• π0 Technical Paper — Architecture details and benchmark results for the foundational generalist policy
• π0.5 Technical Writeup — Open-world generalization results and training methodology