Moonvalley AI

Skild AI is developing a foundation model for robotics - a universal controller designed to operate across different robot morphologies including arms, humanoids, and quadrupeds. Founded in 2023 by Carnegie Mellon professors Deepak Pathak and Abhinav Gupta, the company is building systems that learn from observation and adapt to new robot bodies and environments without requiring task-specific training. The technical approach centers on massive simulation combined with deep learning to create what the company describes as physically-grounded intelligence. The architecture aims to enable robots to generalize across tasks and form factors, analogous to how large language models generalize across text domains. The company raised $300 million in Series A funding and has established partnerships with NVIDIA and Amazon. Target applications include manufacturing and logistics, where the ability to deploy a single model across different robot types could reduce the engineering overhead of task-specific programming. Technical domains span computer vision, deep learning, and robotics, with emphasis on simulation-based training methods that can scale across diverse robotic platforms. Skild AI applies foundation model techniques to the control problem in robotics, tackling the complexity of physical interaction, real-time control constraints, and the reality gap between simulation and deployment. The company is research-driven, focused on moving algorithms from academic settings into physical systems that operate in industrial environments.

We're Gatik, and we're building autonomous trucks that handle the middle-mile logistics for some of North America's biggest retailers. Instead of chasing long-haul trucking or robotaxis, we focused on the regional routes between distribution centers and stores - where goods move frequently but predictably. This is where our Level 4 autonomous system, Gatik Driver™, really shines. It's built specifically for fixed, repeatable routes that retailers depend on every single day, making it possible to remove the driver entirely while keeping operations safe and efficient. What makes us different is that we're not running experiments or small pilots. Our trucks are commercially deployed right now, moving freight for Fortune 50 companies like Walmart, Kroger, and Loblaw across multiple states. We've completed over 60,000 fully driverless deliveries without an incident, and we're the first company in the U.S. operating driverless trucks at commercial scale. Our team combines expertise in autonomous systems with a pragmatic approach to solving real supply chain problems - no hype, just reliable freight that keeps shelves stocked and businesses running smoothly.

In 2025, Zhou Xian and Théophile Gervet recognized a fundamental bottleneck in the robotics field: while digital AI had made extraordinary progress, physical AI - the intelligence enabling machines to perceive, understand, and interact with the real world - lagged dramatically behind. Physical labor contributes an estimated $30-40 trillion to global GDP, yet over 95% remains unautomated because current robotic systems rely on brittle, rigid, and overfitted software stacks that are narrow in scope, costly to deploy, and impossible to scale. They founded Genesis AI to bridge this gap, assembling a team of world-class talent from Mistral AI, NVIDIA, Google, CMU, MIT, Stanford, Columbia, and UMD. Their vision: build a universal robotics foundation model and horizontal robotics platform that would unlock unlimited physical labor. Genesis would achieve this through a scalable data engine fusing real-world robot interaction with high-fidelity physics simulation, an open-source ecosystem to accelerate field-wide progress, and robust robot deployments thriving in messy, unconstrained environments. With $105 million in seed funding from Eclipse Ventures and Khosla Ventures, Genesis emerged from stealth in July 2025 to build what's next in physical AI.

Mistral AI is a French artificial intelligence company founded in April 2023 by Arthur Mensch, Guillaume Lample, and Timothée Lacroix - researchers with prior experience at Google DeepMind and Meta who share academic roots at École Polytechnique. The company develops and releases open-weight generative AI models, positioning them as competitive alternatives to proprietary solutions while maintaining a commitment to transparency and accessibility in frontier AI development. The company's core technical approach centers on training large language models in-house and releasing them with open weights, allowing developers, enterprises, and institutions worldwide to deploy these models while maintaining full control over their data and applications. Mistral AI's models are designed to compete with closed, proprietary systems, and the company frames community-backed development as a methodology for addressing censorship and bias in AI systems. This technical philosophy distinguishes their work from the conventional closed-model paradigm dominant among major AI providers. Mistral AI's business model targets a broad user base spanning software developers, enterprises, and institutions globally. The company positions itself as building a credible alternative to what it characterizes as an emerging AI oligopoly, advocating for open access to frontier AI capabilities rather than concentration of advanced models within a small number of organizations. Their released models are intended to enable external parties to build and operate AI applications independently, without dependency on proprietary API services or restricted model access.

In April 2024, a group of founders recognized a critical gap in modern defense capabilities: the urgent need for AI-driven autonomous systems that could keep pace with the speed and complexity of 21st-century warfare. Traditional defense contractors struggled to deliver attritable, scalable solutions that democratic allies required for operational dominance. From this insight, Harmattan AI was born - a next-generation defense prime dedicated to building the future of autonomous warfare. Starting from a conviction that liberal democracies and their allies deserve superior technological capabilities, the company rapidly developed vertically integrated systems including ISR drones, counter-UAS platforms, and electronic warfare solutions. With $200 million in Series B funding led by Dassault Aviation at a $1.4 billion valuation, Harmattan AI has transformed from a startup vision into a unicorn defense contractor delivering mission-ready systems that enhance operational edge and readiness for allied forces worldwide.

Sanctuary AI, founded in 2018 and headquartered in Vancouver, Canada, develops industrial-grade humanoid robots designed to address labor challenges in manufacturing, logistics, and automotive sectors. The company focuses on embodied AI systems that model human movement and cognitive processes, enabling robots to learn from and interact with physical environments in human-like ways. Sanctuary AI ranks third globally in humanoid robotics patents, reflecting substantial technical development in the field. The company's flagship product, Phoenix, is positioned as a general-purpose humanoid robot built for industrial deployment. Phoenix features human-like dexterity, tactile feedback systems, and fine manipulation capabilities, enabled by a proprietary AI control system. The robot is designed for collaborative operation alongside human workers in automotive plants, manufacturing facilities, and logistics operations, with particular emphasis on tasks characterized as dull, dirty, or dangerous. Sanctuary AI's technical approach centers on several core domains: robotic dexterity and tactile sensing for fine manipulation tasks, proprietary AI control architectures for autonomous operation, and systems designed to achieve general-purpose utility across multiple industrial environments. The company's engineering efforts focus on creating robots that can perform varied tasks within existing industrial infrastructure, rather than requiring specialized environments or fixed automation sequences. Under CEO James Wells, the team develops solutions intended to scale across multiple industry verticals while maintaining industrial-grade reliability standards.