← STARTUPSEMBODIED AI · BASELINE 2026-04-20

Physical Intelligence

confidence: high · 3 watch questions · 7 evidence citations

01 TEAM & SIZE

Founded 2024 (SF). Seven co-founders: Karol Hausman (CEO, ex-Google Brain manipulation), Sergey Levine (Chief Scientist, UC Berkeley), Chelsea Finn, Brian Ichter (VP Eng), Adnan Esmail (hardware), Lachy Groom, Quan Vuong. Team drawn from DeepMind/Google Brain/Stanford/Berkeley. Exact headcount unverified.

02 FUNDING & STAGE

  • Total raised: ~$1.1B (cumulative).
  • Nov 2025: $600M at $5.6B post led by Alphabet's CapitalG; participants incl. Lux, Thrive, Bezos, Index Ventures, T. Rowe Price [S-1-001].
  • Mar 2026: reported talks for ~$1B at ~$11B valuation (roughly 2× in 4 months) [S-1-002].
  • Prior backers: Sequoia, Khosla, Bond, Redpoint, OpenAI, Jeff Bezos.
  • Stage: Series C in progress as of baseline date; no confirmed close.

03 PRODUCT STATE

Foundation-model-for-robots stack, model family all called "π":

  • π0 (Oct 2024): first generalist VLA flow model.
  • π0-FAST: autoregressive variant w/ FAST action tokenizer.
  • π0.5 (Apr 2025): open-world generalization, mobile manipulation in unseen homes, ~10–15-min multi-stage tasks [S-1-003].
  • π*0.6 (Nov 2025): VLA + RL.
  • π0.7 (Apr 2026): "compositional generalization" — matches task-specialist models on coffee-making, laundry-folding, box-assembly without task-specific training [S-1-004].
  • Open-source: Physical-Intelligence/openpi GitHub, Apache-2.0, ~11.4k stars; releases include π0, π0-FAST, π0.5 checkpoints + PyTorch support (Sep 2025) [S-1-005]. π0.7 not yet in repo as of baseline.
  • Hardware embodiments supported: Franka, UR5, xArm, ALOHA bi-manual, DROID, mobile manipulators, humanoid upper-body.

04 GTM MOTION

Research-lab-first + partner-embedded. Blog-driven distribution; open-weights for older generations as top-of-funnel developer mindshare; "collaborations with a number of companies and robotics labs" for hardware + data — specific customers unverified. No public pricing, no self-serve. Positioned as horizontal robot-brain layer, not hardware maker.

05 CORE THESIS

One generalist policy can control many robots across many tasks — bet that a VLA + large cross-embodiment data + RL will reach an LLM-style capability inflection. π0.7's compositional-generalization claim, if robust, is the central proof point.

06 PUBLIC SIGNALS INVENTORY

Bloghttps://www.pi.website/blog
RSSunverified
GitHubhttps://github.com/Physical-Intelligence/openpi (~11.4k stars)
Twitter@physical_int; founders @hausman_k, @svlevine
LinkedIncompany page unverified; Hausman https://linkedin.com/in/karolhausman/
Careershttps://www.pi.website/join-us
YouTubeunverified (demos embedded on blog)
PressTechCrunch, Bloomberg, The Generalist (Hausman interview), CapitalG portfolio post

07 52-WEEK QUESTIONS

  1. Q1
    Will the ~$11B Series C close and be publicly confirmed (filing or press) by Q3 2026?
  2. Q2
    Will π0.7 (or a successor π1.0) be released to `openpi` under Apache-2.0, or will PI shift to closed weights for frontier models?
  3. Q3
    Will PI announce a named commercial customer/deployment (manufacturing, logistics, or home) with revenue or unit-count disclosure before Q2 2027?

08 WEEKLY TIMELINE

0 signals

TIMELINE EMPTY · WEEK 0

Auto-feed fetcher ships Month 2. Manual diffs via Friday script start Week 2.

09 EVIDENCE

7 items · expand
  1. S-1-001
    Bloomberg, "Robotics Startup Physical Intelligence Valued at $5.6 Billion" (2025-11-20) — $600M round led by CapitalG at $5.6B
  2. S-1-002
    Bloomberg, "Ex-DeepMind Staffers' Robotics Startup in Talks for $11 Billion Valuation" (2026-03-27) — ~$1B Series C at ~$11B
  3. S-1-003
    arXiv 2504.16054, "π0.5: a VLA with Open-World Generalization" — mobile manipulation in unseen homes
  4. S-1-004
    TechCrunch, "Physical Intelligence… new robot brain can figure out tasks it was never taught" (2026-04-16) — π0.7 compositional generalization
  5. S-1-005
    GitHub Physical-Intelligence/openpi — Apache-2.0, ~11.4k stars, π0/π0-FAST/π0.5 checkpoints, PyTorch support Sep 2025
  6. S-1-006
    pi.website — mission, product timeline, investors, careers
  7. S-1-007
    The Generalist interview w/ Hausman — team, thesis

10 CONFIDENCE & UNKNOWNS

expand
  • Confidence: high (funding, product, OSS signals all primary-sourced).
  • Unverified: exact headcount; Series C close date/terms; named paying customers; ARR/revenue; RSS feed; company YouTube channel; LinkedIn company URL.
  • Requires outreach: commercial pilot list, robot-embodiment data partners, π0.7 OSS release plans.

11 RELATED

← ALL STARTUPSPHYSICAL AI BRIEFFRONT PAGE