Skild AI

Overview

Skild AI is a robotics AI startup founded by Carnegie Mellon University faculty, developing a general-purpose robot “brain” called Skild Brain. In 2024, they achieved a $1.5B valuation with a $300M Series A round, and in January 2026, reached over $14B valuation with a $1.4B Series C round.

Item	Details
Headquarters	Pittsburgh, PA (additional offices in SF Bay Area and Bengaluru)
Founded	2023
Co-Founders	Deepak Pathak (CEO), Abhinav Gupta (President)
Total Funding	$2.2B+
Valuation	$14B+ (January 2026)
Key Investors	Lightspeed, Coatue, SoftBank Group, Bezos Expeditions, Sequoia, NVIDIA NVentures

Funding History

Round	Date	Amount	Valuation
Seed	2023	$14.5M	-
Series A	July 2024	$300M	$1.5B
Series B	May 2025	$500M	$4.7B
Series C	January 2026	$1.4B	$14B+

Series A investors: Lightspeed Venture Partners, Coatue, SoftBank Group, Bezos Expeditions (Jeff Bezos), Felicis Ventures, Sequoia, Menlo Ventures, General Catalyst, CRV, Amazon, SV Angel, Carnegie Mellon University

Series C investors: SoftBank Group (lead), NVIDIA NVentures, Macquarie Capital, Bezos Expeditions, Disruptive, 1789 Capital, Samsung, LG, Schneider Electric, CommonSpirit, Salesforce Ventures

Founding Team

Deepak Pathak (CEO)

Carnegie Mellon University Robotics Institute Professor
UC Berkeley AI PhD
Expert in self-supervised learning
Research on curiosity-driven reinforcement learning with 4,000+ citations
IIT Computer Science Gold Medalist (2014)

Abhinav Gupta (President)

Carnegie Mellon University Robotics Institute Full Professor (since 2015)
Founding member of Facebook AI Research (FAIR) Robotics
Expert in robotics and computer vision
75,000+ academic citations
NSF, DARPA research funding recipient

Skild Brain

General-Purpose Robot Brain

┌─────────────────────────────────────┐
│            Skild Brain              │
│    "Any Robot, Any Task, One Brain" │
└─────────────────────────────────────┘
          │
    ┌─────┴─────┐
    ▼           ▼
┌───────┐   ┌───────┐   ┌───────┐   ┌───────┐
│Humanoid│   │Quadruped│   │Robot Arm│   │Mobile │
└───────┘   └───────┘   └───────┘   └───────┘

Key Features

Feature	Description
Omni-bodied	Supports all robot forms without specific hardware
Cross-embodiment	Humanoids, quadrupeds, robot arms, mobile manipulators
Hardware Independent	Focus on software models without building proprietary robots
Built-in Safety Constraints	Force output limits when interacting with humans

Training Methodology

Data Sources

Multi-pronged approach to address the scarcity of robot data:

Internet Videos: Learning physics from human behavior
Simulation: Large-scale pre-training using NVIDIA Isaac Lab
World Model: Leveraging NVIDIA Cosmos WFM

Sim-to-Real Transfer

[Internet Videos] + [Simulation] ──Pre-training──→ [Skild Brain] ──Transfer──→ [Real Robot]

Technical Features

Self-Supervised Learning

Utilizing large-scale unlabeled data
Learning physics from internet videos
Curiosity-driven exploration

Scalable Architecture

Applying LLM scaling laws to robotics
More data/compute → Better performance
Cross-robot generalization

Commercial Progress

Revenue Growth

Achieved $0 → $30M revenue within months in 2025
Continued growth in 2026

Application Areas

Security and facility inspection
Last-mile / point-to-point delivery
Warehouses
Manufacturing
Data centers
Construction

Partnerships

LG CNS: Partnership for industrial applications
HPE: Accelerating development with AI solutions

Differentiators

vs Physical Intelligence

Factor	Skild AI	Physical Intelligence
Background	CMU academia	Google DeepMind/Stanford/Berkeley
Founded	2023	2024
Approach	Simulation + video-centric	Real data-centric
Hardware	Software only	Software only
Total Funding	$2.2B+	$1.1B
Valuation	$14B+ (2026.01)	$5.6B (2025.11)
Core Model	Skild Brain	pi-0 (pi-zero)

Simulation vs Real Data Debate

Skild: "Scalable data generation through simulation + internet videos"
      + Leveraging NVIDIA Cosmos/Isaac Lab
      + Overcoming Sim-to-Real gap with Domain Randomization

Physical Intelligence: "Real physical interaction is essential"
                      + Open-sourced pi-0 model (Feb 2025)