Robotics Foundation Models Fundraising (2026)

How robotics foundation model, humanoid, and VLA (vision-language-action) startups raise capital in 2026 after Figure, 1X, Physical Intelligence.

Raising Capital for Robotics Foundation Models & Humanoid Startups

Robotics foundation models are the highest-capital, highest-conviction category of 2025-2026. Physical Intelligence ($400M at $2.4B then $5.6B), Skild AI ($300M at $1.5B+), Figure ($675M at $2.6B then $39.5B rumored), 1X, Apptronik ($350M), Sanctuary AI, Covariant (Amazon licensing exit), and Nvidia GR00T reframed the category. VLA (vision-language-action) models became the architecture consensus. Capital-formation math: $50-500M seed-to-A is now normal for credible teams with DeepMind/OpenAI/Toyota Research/Tesla Optimus lineage.

Why 2026 is different

Figure raised $675M at $2.6B in early 2024, reportedly targeting $39.5B in 2025-2026. Physical Intelligence raised $400M at $2.4B and reportedly $5.6B follow-on. Skild AI raised $300M at $1.5B+. 1X, Apptronik, Sanctuary, and Agility Robotics all raised material rounds. Nvidia GR00T (foundation model for humanoids) and Isaac platform reframed the tooling stack. VLA architectures (Physical Intelligence π0/π0.5, Google RT-2, Nvidia GR00T N1) became the consensus. BMW, Mercedes, GXO, and Amazon reference deployments legitimized commercial timelines. China (Unitree, Xpeng, Xiaomi CyberOne, UBTech) accelerated humanoid competition. Tesla Optimus/Elon reset expectations for cost curves.

Realistic capital stack

Seed: $10-50M with founders from DeepMind, OpenAI, Tesla Optimus, Boston Dynamics, Toyota Research, Google Brain Robotics, Nvidia, or CMU/Stanford robotics. Series A: $75-300M with commercial pilots and demonstrated VLA capabilities. Series B: $300M-$1B at $50M-$1B post. Reference points 2024-2026: Figure ($675M at $2.6B, later ~$39.5B rumored), Physical Intelligence ($400M+ at $2.4B then $5.6B), Skild ($300M at $1.5B+), 1X ($100M B), Apptronik ($350M A at $2.4B rumored), Agility ($150M+), Covariant (Amazon licensing), Sanctuary, Chef Robotics, Dexterity ($95M+).

Common failure modes

Founders without top-tier robotics research lineage — cannot raise at any price. No data-collection or teleop strategy — VLA models require millions of hours of data. Underestimating capex — humanoid programs need $200M+ over 5 years to reach commercial pilots. Selling robots outright instead of RaaS — caps multiples. Ignoring safety certification and standards work. Positioning as 'humanoid for everything' without a specific commercial wedge (auto assembly, logistics, hospitality).

Frequently asked questions

Can new humanoid entrants still raise in 2026?
Only with exceptional research pedigree (DeepMind/OpenAI/Tesla Optimus/Boston Dynamics) or a differentiated data/embodiment thesis. The field has ~8 well-capitalized players; new entrants need a specific wedge or a Chinese-scale supply-chain advantage.
VLA foundation model vs full-stack humanoid — which is fundable?
Both, but they attract different capital. Foundation-model plays (Physical Intelligence, Skild) raise at software-multiple economics on horizontal model licensing. Full-stack humanoids raise on RaaS + strategic value; capex is 5-10x higher.
Realistic exit?
Strategic acquisition by Nvidia, Tesla, Toyota, Hyundai, Samsung, or a Chinese OEM. IPO for category leaders (Symbotic's SPAC path is the recent public comp). Covariant → Amazon (licensing, ~$400M) is the recent reference comp for VLA-model licensing exits.

Related fundraising verticals (40)

Investor directory · Fundraising library · Articles A–Z · Company funding database