Genesis AI raises $500M at $3.5B valuation to build robots in the US
Genesis AI has raised $500M at a $3.5B valuation to build general-purpose robots with dexterity close to that of a human hand. The company is developing everything at once: the AI model that controls the robots, the machines themselves, and individual components such as hand joints. It promises the first fully American-assembled robots next year. For the market, this is a signal: investors are willing to pay for the combination of model and hardware, not just software.
AI-processed from Bloomberg Tech; edited by Hamidun News
Robots that can do more than haul boxes around a warehouse and instead work with their hands almost like a person remain one of the most expensive and difficult problems in AI. Genesis AI has just raised $500M for it. The startup's valuation after the round is $3.5B. The company is building general-purpose robots that, over time, should perform complex tasks with human-like dexterity. It plans to release its first fully American-made robots next year.
What is known about the round and the plans
CEO and co-founder Zhou Xian described this on Bloomberg Tech. He named three near-term goals: scale the robotic AI model, release the first robots built entirely in the US, and then ramp up production.
The word "general-purpose" matters here. Most commercial robots today are narrow: one manipulator performs one operation on one line. A general-purpose robot should handle different tasks without being fully reprogrammed. That requires a model that transfers a skill from one object and situation to others.
The size of the round shows that investors are willing to pay for a long-term bet on physical AI. No figures on revenue, customers or unit volumes were mentioned in the interview, so the $3.5B valuation currently rests on technology and team rather than confirmed sales.
Why build everything in-house
Zhou Xian also explained why Genesis chose the full stack: the company works on AI, the robots themselves and their individual components. In a market where many startups build only the "brain" and buy the body from suppliers, this is a notable choice.
The logic is easy to picture with an everyday example. A cook performs well when the knife is balanced for their hand. If the knife is awkward, heavy or poorly transmits pressure, even a skilled cook cuts worse. The same goes for a robot. The model learns from signals coming from sensors, grippers and joints. If the hardware is noisy, has play or poorly senses touch, the model has to compensate for weak parts, and dexterity does not improve. When one team designs both the model and the hand, it can tune them to each other: change the sensors to suit training needs and teach the model with the real mechanical constraints in mind.
There is also a production argument. Manufacturing in the US is usually more expensive than in Asia, but it gives control over supply and fast iteration: engineers can take apart, rework and reassemble a unit themselves without waiting for a container from overseas.
Risks and what matters for practitioners
The limitations are also clear. Human-like dexterity is still a goal, not a result: the source says the robots should "ultimately" perform complex tasks. A full stack means spending on three expensive areas at once: AI research, hardware engineering and a factory. A mistake in any of them will slow the others. In addition, scaling from first prototypes to series production usually runs into the reliability and cost of components.
What this means for readers:
- For founders: the physical AI market is attracting large sums again, and vertical integration looks like a workable strategy rather than an overreach.
- For product managers and engineers: watch whether Genesis gets pilot customers and publishes open data on task success rates. Those will show how close the claimed dexterity is to reality.
- For robotics developers: the combination of "model plus proprietary hardware" may set a benchmark against which off-the-shelf platforms will have to be compared.
The next checkpoint comes in 2027. If Genesis ships its first American-built robots on the promised timeline and shows they handle real tasks, the $3.5B valuation will get its first tangible confirmation.
- Nov 21, 2026Genesis AI Raises $500M at $3.5B Valuation to Teach Robots Through Hand Movements
- Nov 21, 2026Paris-Based Genesis AI Raises $500M at $3.5B Valuation for Eno Robots
- Jul 24, 2026UK Startup Humanoid Raises $152M Series A at $1.35B Valuation
- Jun 1, 2026Genesis AI releases Genesis World 1.0, a platform that evaluates robots 400 times faster
Key Questions Answered
What is known about the round and the plans?
CEO and co-founder Zhou Xian described this on Bloomberg Tech. He named three near-term goals: scale the robotic AI model, release the first robots built entirely in the US, and then ramp up production. The word "general-purpose" matters here. Most commercial robots today are narrow: one manipulator performs one operation on one line. A general-purpose robot should handle different tasks without being fully reprogrammed. That requires a model that transfers a skill from one object and situation to others.
Why build everything in-house?
Zhou Xian also explained why Genesis chose the full stack: the company works on AI, the robots themselves and their individual components. In a market where many startups build only the "brain" and buy the body from suppliers, this is a notable choice. The logic is easy to picture with an everyday example. A cook performs well when the knife is balanced for their hand. If the knife is awkward, heavy or poorly transmits pressure, even a skilled cook cuts worse. The same goes for a robot.
What are the key details on Risks and what matters for practitioners?
The limitations are also clear. Human-like dexterity is still a goal, not a result: the source says the robots should "ultimately" perform complex tasks. A full stack means spending on three expensive areas at once: AI research, hardware engineering and a factory. A mistake in any of them will slow the others. In addition, scaling from first prototypes to series production usually runs into the reliability and cost of components.
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.