Automation

Genesis AI raises $500M Series A for dexterous humanoid robots

The Carnegie Mellon spinout uses sensor gloves and force data to train robots for complex manipulation tasks like packaging and wire handling.

Omega Editorial · November 17, 2026 · 3 min read

Key takeaways

  • Genesis AI raised $500 million in Series A funding led by Khosla Ventures, Premji Invest, and Eclipse to scale its dexterous humanoid robot platform.
  • The company collects force data from 100+ human workers wearing sensor gloves to train robots for complex manipulation tasks like packaging and wire handling.
  • Genesis has built 30-40 units of its Eno humanoid robot and selected a U.S. manufacturing partner for mass production starting next year.
  • The startup's dexterous robotic hands currently require replacement after only a few weeks of regular operation, highlighting ongoing hardware durability challenges.
  • CEO Zhou Xian argues that training infrastructure and data quality matter more than AI model architecture, which he estimates contributes only 10% to performance.

Genesis AI has closed a $500 million Series A funding round to advance its approach to robotic dexterity, the startup announced this week. Khosla Ventures, Premji Invest, and Eclipse led the investment in the company, which is building humanoid robots designed to handle complex manipulation tasks.

The Silicon Valley-based startup, founded in 2024 by Carnegie Mellon graduates Zhou Xian and Théophile Gervet, previously raised a $105 million seed round in July 2025. The company's core thesis centers on what CEO Xian calls "cracked dexterous manipulation" — the ability for robots to perform intricate hand movements that current automation struggles to complete reliably.

The glove-based training approach

Genesis differentiates itself through its data collection method. The company equips more than 100 human workers worldwide with sensor-enabled gloves and headset cameras. These workers perform tasks while the gloves capture force data — what Genesis considers the critical unlock for effective manipulation.

The startup's robots wear identical sensor-equipped gloves, and their hands are engineered to match human hand proportions. This design fidelity enables capabilities in sorting and wire manipulation, according to Xian.

During a demonstration for TechCrunch, Genesis showed off its physical AI model operating two off-the-shelf robotic arms. The system picked up small cardboard boxes, scanned them, placed them in a larger box, folded and taped it closed, then applied an adhesive label. In a second run, the robot fumbled a box and struggled with the label — what a Genesis executive termed "hiccup data" that can improve the model.

Hardware remains the bottleneck

Genesis is pursuing vertical integration, generating its own training data, developing its own generative AI model, and designing its own humanoid robot called Eno. The company has built 30 to 40 Eno units and selected a mass-manufacturing contractor to begin U.S. production next year, though it declined to share production targets.

Eno wasn't demonstrated during TechCrunch's visit. The robot, designed with aesthetic sensibilities that recall Apple products rather than industrial machinery, is being piloted with undisclosed customers.

Xian identified hardware capability as the key constraint, particularly in mechanical design and supply chains for industrial-grade components. The company's dexterous hands currently require replacement after just a few weeks of regular operation.

On the software side, Xian downplayed architectural distinctions between different AI model types, arguing that training infrastructure and data quality matter more than model architecture, which he estimates contributes only 10% to final results. The company uses physics simulations — developed with adviser Minchen Li, co-author of an influential physics algorithm — to generate millions of high-fidelity training scenarios.

Why it matters

The robotics industry faces a critical test: whether AI-powered humanoid robots can bridge the gap between impressive lab demonstrations and reliable industrial deployment. Genesis's $500 million raise reflects investor confidence that force-sensing data and simulation can solve the dexterity problem that has limited warehouse and manufacturing automation. But the company's own hardware durability challenges — hands lasting only weeks — illustrate the engineering hurdles that remain before humanoid robots can compete economically with human workers or specialized automation.

The funding and technical details were first reported by TechCrunch.

  • #humanoid robots
  • #robotics funding
  • #dexterous manipulation
  • #physical ai
  • #warehouse automation
  • #khosla ventures

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Automation