Genesis AI emerged from stealth in July 2025 with an ambitious mission: to unlock the automation of physical labour at scale through a universal robotics foundation model. The company’s $105 million seed round, co-led by Eclipse Ventures and Khosla Ventures, and joined by Bpifrance, HSG, Eric Schmidt, Xavier Niel, and HongShan, instantly positioned Genesis AI as a key contender in the rapidly evolving robotics AI sector.
The problem Genesis AI set out to solve is both vast and urgent. Despite decades of technological progress, most physical labour worldwide, spanning logistics, manufacturing, healthcare, agriculture, and domestic work, remains unautomated. Traditional robotics systems are limited by their narrow specialisation, often coded for single repetitive tasks and unable to generalise across new environments or hardware.
The bottleneck is not just in hardware but data: training robots requires immense amounts of real-world data, which is costly, slow, and often impractical to collect at scale. Existing simulation tools, meanwhile, are either too slow or lack the fidelity needed to produce robust, generalizable AI for robotics.
Genesis AI’s solution is to build a general-purpose robotics foundation model trained not on text or images but on simulations of the physical world. At the heart of this approach is a proprietary, ultra-fast physics simulation engine, originating from an academic collaboration across 18 universities and led by co-founder Zhou Xian. This engine can generate high-fidelity synthetic data at speeds up to 430,000 times faster than real-world time.
Founders’ background, mission, and vision
Genesis AI was founded in December 2024 by Zhou Xian, a Carnegie Mellon PhD in robotics, and Théophile Gervet, a former research scientist at French AI lab Mistral and a PhD in AI from Carnegie Mellon. The founders’ backgrounds reflect a blend of deep academic rigour and practical experience at the cutting edge of AI and robotics.
Their mission is to unlock unlimited physical labour by building robust, flexible, and cost-efficient general-purpose robotics powered by physical AI. They envision a world where robots, powered by general-purpose AI, seamlessly integrate into every aspect of society, boosting productivity, safety, and quality of life.
The motivation for starting Genesis AI stemmed from their firsthand experience with the limitations of existing simulation tools and the lack of a collaborative, open ecosystem in robotics research. They saw the need for a user-friendly, transparent, and collaborative platform to democratize access to high-quality simulation data and accelerate the development of embodied intelligence.
Notably, Genesis AI’s founding team includes over 20 researchers from top institutions such as MIT, Stanford, Columbia, and UMD and experience from leading tech companies including Nvidia and Google. This diverse expertise strengthens the company’s ability to tackle the complex challenges of robotics AI.
Synthetic data as a training shortcut
Technologically, Genesis AI’s edge lies in its proprietary simulation engine and its commitment to a closed-loop data generation process that combines synthetic and real-world data. This approach creates a robust training loop, improving model performance and generalizability far beyond what is possible with real-world data alone. The company’s foundation model architecture is designed to generalise across multiple hardware types and tasks, mirroring the success of large language models in text-based domains.
Genesis’s simulation tools are significantly faster and more cost-effective than industry standards like Nvidia’s Isaac Gym and offer greater data diversity and quality. Unlike competitors such as Skild AI and Physical Intelligence, which rely more heavily on off-the-shelf simulators and closed platforms, Genesis is committed to open collaboration. It plans to open-source components of its simulation engine and foundation model to foster global research and innovation.
A key differentiator is Genesis AI’s dual headquarters in Silicon Valley and Paris, enabling access to American and European talent pools and markets. The company’s closed-loop system for synthetic and real data, combined with its open-source philosophy, sets it apart in speed, flexibility, and ecosystem development.
Investor confidence in Genesis AI
The recent funding will be used to scale synthetic data infrastructure and expand the team of robotics, machine learning, and computer graphics specialists. Genesis is committed to building an open, user-friendly platform for robotics research and development, lowering barriers to entry for innovators worldwide.
In addition, Genesis AI is developing partnerships with leading academic and industry labs to further validate and extend its technology across new domains. The long-term vision is for Genesis AI to become the backbone of a new era of automation, enabling robots to learn and adapt in virtual worlds before seamlessly operating in the real world, ultimately transforming the $40 trillion global market for physical labour.
Kanu Gulati, partner at Khosla Ventures, said, “Of all the teams we’ve seen, we like Genesis’s approach to robotics foundation models. Whether anyone will be able to build one that generalises across tasks is a big unknown – that’s the bet we want to go after.”
Genesis joins a fast-growing cohort of startups targeting general-purpose robotic intelligence. Competitors in the space include Skild AI, reportedly valued at $4 billion earlier this year, and Physical Intelligence, which raised a $400 million round.
Next steps
Genesis AI plans to release an early version of its foundation model to the broader robotics research community by the end of 2025, aiming to foster transparency and accelerate progress across the field. The company’s platform is designed to support robotic systems in various industries, including logistics, manufacturing, agriculture, healthcare, and domestic services, where automation can have the greatest impact.
The announcement positions Genesis as a key player in the evolving interface between AI and real-world automation. The combination of physics-based simulation and foundation model architecture is expected to unlock transformative new use cases.
