Genesis AI Raises $105M to Build AI Brains for Robots

Post by : Amit

A new contender in the race to power intelligent machines has just emerged from stealth. Genesis AI, a Silicon Valley-based startup, has announced its launch today with a staggering $105 million seed round—one of the largest early-stage fundings in AI history. The company’s bold mission: to build the next-generation foundational AI models tailored specifically for robots.

Led by co-founders Sam Fok and Mikhail Svyatoslavsky, both alumni of MIT’s robotics and applied AI labs, Genesis AI is not chasing general-purpose language models like OpenAI or Anthropic. Instead, it wants to become the cognitive engine inside physical machines—from warehouse bots and surgical arms to drones and humanoids.
“We believe the future of AI isn’t just text and vision. It’s embodiment,” said Fok. “Machines need models built for motion, manipulation, and the messy physical world.”

The startup’s jaw-dropping $105M seed was co-led by Eclipse Ventures and Khosla Ventures, with major participation from Sequoia Capital, NVIDIA Ventures, and Toyota Ventures. The funding reflects growing investor appetite for robot-native AI models as industries scramble to automate logistics, manufacturing, eldercare, and beyond.

Eclipse, known for its bets on industrial hardware and deep tech, said Genesis AI is “the software platform robotics has been waiting for.” Notably, the startup has also secured early access partnerships with top robotics OEMs in the U.S., Japan, and South Korea.

While traditional large language models (LLMs) are optimized for words, Genesis AI is developing what it calls “Motion Foundation Models” (MFMs)—AI systems trained on millions of hours of sensorimotor data. This includes force feedback, video, proprioception, haptic responses, and mechanical telemetry—collected from robotic arms, autonomous vehicles, and embedded sensors.

Its models aim to:

• Enable zero-shot task execution in physical environments
• Integrate seamlessly with robot operating systems (ROS2 and beyond)
• Learn dexterity, timing, and coordination without retraining per task

The ultimate goal is to allow a robot to watch a human perform a task once—and replicate it across environments, with contextual adaptability.
“We’re building the GPT of grasping, the Claude of contact,” quipped Svyatoslavsky. “Every future robot should be born with a world-ready brain.”

Genesis AI’s launch signals a new era in robotics—one in which robots are no longer just programmable tools, but AI-native agents that reason, adapt, and even collaborate.

Industry analysts see this as a pivotal shift. While OpenAI’s ChatGPT, Google DeepMind’s Gemini, and Meta’s LLaMA models dominate headlines, none are designed to control hardware at scale. Genesis fills a void where embodied intelligence meets real-world application.

Early applications will target:

• Warehouse and logistics automation (in partnership with AMR manufacturers)
• Surgical robotics with real-time precision control
• Agricultural robots for harvesting and soil analysis
• Dexterous assembly-line co-bots

The launch of Genesis AI reignites debate over the role of embodiment in Artificial General Intelligence (AGI). While most AGI discussions focus on language and logic, a growing school of thought argues that true intelligence requires physical grounding.

The company remains in stealth mode about its upcoming products, but teased that its first public demo, scheduled for late 2025, will feature a robotic system completing 20+ complex tasks in a live environment, without code or teleoperation.

AI, Robotics