Skild AI

Market position

Sacra positions Skild among horizontal robot foundation model providers competing with Covariant, Physical Intelligence and Genesis AI, and against big-tech platforms (Alphabet's Intrinsic, NVIDIA Isaac) and vertically integrated humanoid players (Figure AI, Tesla Optimus, Apptronik, Galbot) that build proprietary data flywheels. TechCrunch situates Skild alongside Field AI and 1X in the push toward adaptable, learn-as-you-go robotic software amid humanoid hype. Forbes describes it as a $14 billion-valued startup with roughly $3 billion in venture funding.

Skild frames its differentiation as generality: unlike models overfit to a specific robot type, the Skild Brain is described as omni-bodied and able to control robots even without knowing their exact body, adapting via in-context learning. NVIDIA cites adaptation to low-cost, noisy hardware ($4,000-$15,000 systems versus $250,000+ customized systems), robustness to mechanical failures, and reliance on scalable synthetic and human-video data instead of expensive teleoperated collection. Cofounders claim to Forbes that the model is trained on 1,000 times more data than rivals.

Technology

The Skild Brain is trained largely on non-teleoperated data: physics-based synthetic data generated in simulation and human videos from the internet, from which the company extracts affordances by treating humans as "biological robots." NVIDIA's case study states Skild uses NVIDIA accelerated computing, Omniverse libraries, Isaac Lab for physics simulation and reinforcement learning, and Cosmos Transfer to augment datasets with varied environments, lighting and visual features. Skild runs massive parallel simulations with thousands of robot instances across humanoid, quadruped and arm embodiments in thousands of environments, which the company says lets it acquire "a millennium of experience" in days and forces in-context learning strategies rather than hardware-specific memorization. NVIDIA reports the resulting model adapts to new embodiments with zero or minimal post-training and exhibits zero-shot generalization such as walking on stilts. Sacra describes training on trillions of synthetic episodes plus millions of real-world video frames, followed by fine-tuning from Skild and partner deployments.

Go-to-market

Coverage describes a platform/API distribution approach: robot specifications are uploaded to Skild Cloud, control interfaces are auto-generated, and customers call high-level APIs or issue natural-language instructions (Sacra). Skild also develops first-party robot application platforms (security/inspection, mobile manipulation, autonomous packing) to demonstrate real-world value, and works with partners including NVIDIA, which both invests in and deploys Skild technology.

Robotics OEMs and enterprise buyers of automation, according to Sacra. Application focus areas cited by the company and its coverage include industrial manufacturing, warehousing, construction, healthcare, and security/inspection work in dangerous environments such as oil rigs and machine rooms, framed against US labor shortages.

Geography

Headquartered in Pittsburgh, Pennsylvania, United States (Sacra, Contrary Research, Forbes). Coverage references US deployments, including NVIDIA's Houston, Texas factory.

History

Founded in 2023 and operating in stealth, Skild AI raised a $14.5 million seed round in 2023 co-led by Lightspeed and Sequoia (per Sacra). It emerged from stealth on 9 July 2024 with a $300 million Series A at a $1.5 billion valuation led by Lightspeed, Coatue, SoftBank Group and Bezos Expeditions, with participation from Felicis, Sequoia, Menlo Ventures, General Catalyst, CRV, SV Angel, Carnegie Mellon University, and Amazon's Industrial Innovation Fund and Alexa Fund. A further round at a reported $4.5 billion valuation followed in mid-2025 (rumored around $500 million, amount undisclosed). In January 2026 TechCrunch reported a $1.4 billion Series C led by SoftBank at more than $14 billion, with Nvidia, Macquarie Group and 1789 Capital participating. Sacra separately reports a $1.4 billion round announced in March 2026 at a $15 billion valuation and approximately $1.7 billion in total funding, while Forbes reports about $3 billion raised as of April 2026.

Risks & controversies

Sacra's competitive analysis flags structural risks rather than specific controversies: big-tech platforms (NVIDIA Isaac, Alphabet's Intrinsic) could bundle robotic intelligence with hardware or enterprise software and subsidize development, creating pricing pressure on pure-play robotics AI vendors; vertically integrated players such as Figure AI, Tesla and Apptronik generate proprietary data flywheels from commercial deployments; open-source strategies such as Physical Intelligence's could commoditize the foundation-model layer; and China-based Galbot benefits from component cost advantages of 40-60%. If large vertically integrated players build their own foundation models, the addressable market for horizontal providers like Skild could shrink. Sources also disagree on funding details: TechCrunch dates the $1.4 billion round to January 2026 at a $14 billion valuation while Sacra dates it to March 2026 at $15 billion, and total-funding figures range from about $1.7 billion (Sacra) to about $3 billion (Forbes).