Will Nvidia and Samsung's Skild AI Investment Reshape Robotics AI?

Nvidia and Samsung are planning strategic investments in Skild AI, the Pittsburgh-based startup developing foundation models for robotic control, according to sources familiar with the discussions. The dual investment signals intensifying competition in the Physical AI space, where general-purpose models promise to enable robots to perform complex tasks without task-specific programming.

Skild AI, founded in 2023 by former Carnegie Mellon researchers, has raised $300 million across multiple rounds while developing what it calls "universal robot intelligence." The company's approach centers on training large foundation models using massive datasets of robot interactions, enabling zero-shot generalization across different robot platforms and tasks. This contrasts sharply with traditional robotics approaches that require extensive hand-engineering for each specific application.

The investment timing appears strategic for both tech giants. Nvidia's interest aligns with its broader AI infrastructure play, potentially integrating Skild's models with its Omniverse simulation platform and Jetson edge computing hardware. Samsung, meanwhile, sees robotics as a key growth vector beyond consumer electronics, with applications spanning manufacturing automation to home assistance robots.

How Skild AI's Technology Works

Skild AI's core innovation lies in training foundation models on diverse robotic datasets, similar to how large language models learn from text corpora. The company collects teleoperation data, simulation runs, and real-world robot interactions to build what CEO Abhinav Gupta describes as "the ImageNet for robotics."

Their models demonstrate particular strength in dexterous manipulation tasks, where traditional programmatic approaches struggle with the complexity of hand-object interactions. Early demonstrations show robots performing tasks like folding laundry, assembling electronics, and manipulating deformable objects—capabilities that typically require months of specialized development.

The foundation model approach offers compelling advantages over traditional robotics software stacks. Rather than programming specific behaviors, engineers can prompt Skild's models with natural language instructions or visual demonstrations. This dramatically reduces development time while enabling robots to adapt to novel scenarios they haven't explicitly encountered during training.

Strategic Implications for Nvidia

Nvidia's investment interest reflects the company's broader push into embodied AI, following its GR00T humanoid robot foundation model announcement in 2024. By backing Skild AI, Nvidia positions itself as the infrastructure provider for the entire robotics AI stack—from training hardware to inference chips to foundational models.

The partnership could accelerate Skild's development timeline while creating a powerful feedback loop. Nvidia's superior hardware capabilities would enhance Skild's model training, while Skild's software would drive demand for Nvidia's robotics hardware ecosystem. This vertical integration strategy mirrors Nvidia's successful approach in autonomous vehicles and datacenter AI.

Industry analysts note that Nvidia has been aggressively courting robotics AI startups, with investments in Physical Intelligence (π) and partnerships with companies like Figure AI. The Skild AI investment would further cement Nvidia's role as the dominant platform provider for next-generation robotics.

Samsung's Robotics Ambitions

Samsung's involvement reflects the Korean conglomerate's serious robotics investments across multiple business units. The company has deployed thousands of robots in its semiconductor fabs and smartphone assembly lines, giving it unique insights into industrial robotics pain points.

Beyond manufacturing, Samsung sees consumer robotics as a potential $50 billion market opportunity by 2030. The company's vision includes household robots that can perform complex tasks like cooking, cleaning, and eldercare assistance. Skild AI's foundation models could accelerate Samsung's timeline for deploying such systems commercially.

Samsung's investment would also provide Skild AI with crucial manufacturing and deployment expertise. Unlike pure software companies, Samsung understands the challenges of scaling robotic systems from laboratory demonstrations to mass production—knowledge that proves invaluable for robotics startups.

Market Context and Competition

The investment comes as the robotics foundation model space intensifies rapidly. Physical Intelligence (π) recently raised $400 million at a $2.4 billion valuation, while Tesla continues developing its own foundation models for Optimus. OpenAI has also hinted at robotics ambitions, though specific timelines remain unclear.

Skild AI's approach differs from competitors through its emphasis on multi-modal learning across diverse robot platforms. While companies like Physical Intelligence focus primarily on manipulation tasks, Skild trains models that handle both locomotion and manipulation—crucial for humanoid applications.

The funding environment for robotics AI remains robust despite broader tech market volatility. Investors recognize that foundation models represent a paradigm shift similar to the transition from rule-based AI to deep learning in computer vision and natural language processing.

Technical Challenges Ahead

Despite promising early results, Skild AI faces significant technical hurdles. Foundation models for robotics must handle physical constraints, safety requirements, and real-time control loops that don't exist in purely digital applications. The sim-to-real gap remains a fundamental challenge, with simulation-trained models often struggling in messy real-world environments.

Safety represents another critical concern. Unlike language models that generate text, robotic foundation models control physical systems capable of causing harm. Ensuring reliable behavior across diverse scenarios requires extensive validation that traditional software testing doesn't address.

The compute requirements for training robotics foundation models also dwarf those of language models. Physical interactions generate massive amounts of sensory data, while the need for real-time inference constrains model architectures in ways that batch processing applications don't face.

Key Takeaways

  • Nvidia and Samsung are planning strategic investments in robotics foundation model startup Skild AI
  • The investment reflects intensifying competition in Physical AI, with foundation models promising to replace task-specific robot programming
  • Skild AI differentiates through multi-modal training across diverse robot platforms, enabling both locomotion and manipulation
  • Nvidia's investment aligns with its broader embodied AI strategy, creating a vertical integration opportunity from chips to software
  • Samsung brings crucial manufacturing and deployment expertise that could accelerate Skild's commercial timeline
  • Technical challenges remain significant, particularly around sim-to-real transfer, safety validation, and compute requirements

Frequently Asked Questions

What makes Skild AI different from other robotics AI companies? Skild AI focuses on training foundation models using diverse robotic datasets across multiple platforms, enabling zero-shot generalization to new tasks and robots. This contrasts with companies that develop task-specific or robot-specific solutions.

How much funding has Skild AI raised total? Skild AI has raised $300 million across multiple funding rounds since its founding in 2023, with the Nvidia and Samsung investments representing additional strategic capital.

When will Skild AI's technology be commercially available? While Skild AI hasn't announced specific commercial timelines, the company is currently working with select partners on pilot deployments. Broader commercial availability likely depends on continued model development and safety validation.

How does Skild AI's approach compare to Tesla's Optimus development? Tesla develops its own foundation models specifically for Optimus humanoids, while Skild AI creates general-purpose models that work across different robot platforms. Tesla focuses on vertical integration while Skild provides horizontal AI infrastructure.

What are the main technical risks for robotics foundation models? Key risks include sim-to-real transfer gaps, safety validation challenges, and the massive compute requirements for training and inference. Physical constraints and real-time control requirements also add complexity not present in digital AI applications.