Skild AI

Skild AI raised $1.4 billion in a funding round announced in March 2026, valuing the company at $15 billion. The company previously raised $300 million in a Series A in July 2024, led by Lightspeed Venture Partners, Coatue, SoftBank Group, and Bezos Expeditions, and a $14.5 million seed round in 2023 co-led by Lightspeed and Sequoia Capital, bringing total funding to approximately $1.7 billion.

The broader investor roster includes Sequoia Capital, Felicis Ventures, Menlo Ventures, General Catalyst, CRV, SV Angel, Carnegie Mellon University, Amazon Industrial Innovation Fund, and Alexa Fund.

Skild AI is building a universal robotic foundation model called Skild Brain that acts as plug-and-play intelligence for any type of robot. The system works like an operating system for physical machines—once a robot's joints, sensors, and cameras are mapped to the Brain's API, the robot can perform complex tasks like grasping objects, navigating environments, climbing stairs, or executing pick-and-place operations without task-specific programming.

The architecture uses hierarchical control where a high-level policy decides on broad actions like walk to shelf, grasp box while a low-level policy translates those intentions into precise joint movements executed in milliseconds. The foundation model is trained on trillions of synthetic episodes in simulation plus millions of real-world video frames, then fine-tuned with targeted robot deployments from Skild and partners.

Robot manufacturers and enterprises upload their robot specifications to Skild Cloud, where the compiler automatically generates the control interface. Users can then call high-level API endpoints or provide natural language instructions. The system works across different robot form factors—humanoids, quadrupeds, warehouse mobile robots, and tabletop arms—through an abstraction layer that requires only small calibration datasets for new hardware. Real-world performance data flows back to improve the global model for all customers, creating a horizontal learning flywheel.

Skild AI operates a B2B software-as-a-service model targeting robotics OEMs and enterprise customers who need intelligent automation without building AI capabilities in-house. The company positions itself as infrastructure for the robotics industry, similar to how cloud platforms serve software companies.

The business model centers on licensing the Skild Brain foundation model through cloud-based APIs, with customers paying for robot intelligence rather than developing proprietary control systems. This approach allows hardware manufacturers to focus on mechanical engineering while outsourcing the AI layer to Skild's platform.

Revenue streams include foundation model licensing, specialized software modules for different verticals like security and inspection, and cloud infrastructure services for training and inference. The company is developing an AI-factory offering that packages training and inference as private cloud services, moving beyond pure software licensing toward managed AI infrastructure.

The model creates network effects as more robot deployments generate training data that improves the foundation model for all customers. This horizontal approach contrasts with traditional robotics software that requires custom development for each use case, potentially creating significant cost advantages and faster deployment cycles.

Covariant leads the pack with six years of proprietary retail data and claims 99% pick accuracy in warehouse environments. Their Composer SDK lets integrators fine-tune tasks without coding, directly challenging Skild's low-code positioning. Genesis AI emerged in July 2025 with $105 million in seed funding, building synthetic physics engines they claim train 3x faster than existing platforms. Physical Intelligence raised $400 million to build PI-1, pursuing an open-source base model strategy that could commoditize the foundation layer where Skild competes.

These competitors are racing to establish the dominant middleware layer for robotics, similar to how operating systems emerged in computing. The winner will likely be determined by training data quality, inference speed, and ecosystem adoption rather than pure technical capabilities.

Alphabet's Intrinsic and NVIDIA's Isaac platform leverage massive compute resources and existing enterprise relationships to build robotic intelligence stacks. NVIDIA's partnership network gives them distribution advantages through hardware vendors, while Intrinsic benefits from Google's AI research and cloud infrastructure.

These platforms pose existential risks to independent players like Skild by potentially bundling robotic intelligence with existing enterprise software or hardware purchases. Their ability to subsidize development costs through other business lines creates pricing pressure for pure-play robotics AI companies.

Figure, Tesla Optimus, and Apptronik are building proprietary AI systems tightly coupled with their hardware platforms. This vertical integration allows them to optimize the full stack but limits their addressable market to their own robots.

The success of vertical players could fragment the market and reduce demand for horizontal platforms like Skild Brain. If hardware manufacturers follow Tesla's playbook of developing proprietary AI, the market for third-party robotic intelligence could shrink significantly.

Skild's foundation model architecture enables expansion into specialized robotics applications beyond general manipulation and navigation. The company is developing vertical-specific modules for construction, healthcare, and security that leverage the same underlying intelligence but add domain expertise.

The mobile manipulation SDK and security inspection stack represent early examples of how Skild can package their foundation model into sellable software products. Each vertical application opens multi-billion dollar market segments while leveraging shared AI infrastructure, creating operational leverage as the company scales.

Initial focus on R&D labs and tech companies is expanding toward industries facing acute labor shortages—manufacturing, warehousing, construction, and healthcare. Each sector represents massive addressable markets with different deployment patterns and willingness to pay for automation solutions.

Strategic investor relationships with SoftBank, Samsung, and Amazon provide direct access to hardware ecosystems and enterprise customers. These partnerships could accelerate adoption by embedding Skild Brain into existing robotic deployments and supply chains, creating distribution advantages over pure-play competitors.

International investors provide natural expansion paths into key robotics markets. SoftBank and Samsung offer access to Japan and Korea, which deploy over 60% of the world's industrial robots. The HPE partnership adds European data center infrastructure, enabling local AI training and inference to meet regulatory requirements.

Regional expansion faces competition from local players who understand specific market dynamics and regulatory environments. However, multinational customers increasingly prefer consolidated vendors over managing multiple regional relationships, potentially favoring Skild's global platform approach.

Model commoditization: Open-source initiatives like Physical Intelligence's PI-1 could commoditize the foundation model layer where Skild competes, forcing differentiation into higher-level tooling and services where margins may be lower and competitive moats weaker.

Hardware fragmentation: The robotics industry remains highly fragmented with proprietary hardware platforms and communication protocols, potentially limiting the addressable market for horizontal AI solutions as manufacturers pursue vertical integration strategies similar to Tesla's approach.

Compute economics: Training and running large robotic foundation models requires massive computational resources, creating ongoing infrastructure costs that scale with customer usage and potentially constraining unit economics compared to traditional software businesses.