Skip to content
HomeBriefFor youTrackersTopicsCompaniesGlossaryNewsletterSecurityHealth & medicineFinance & marketsRoboticsModels & researchPolicy & regulationEnterprise & businessDeveloper tools
Robotics · Nov 18, 2026 · 3 min read

Genesis AI raises $500 million at $3.5 billion for a robot trained on glove data from 100 workers, and GlobalFoundries becomes Atlas's first non-automotive tester. Two routes into the factory, priced very differently

One company is being valued on how it collects hand data before it has customers; the other is earning a humanoid pilot after 90,000 inspections by a four-legged robot. The second route is the one most buyers should copy.

Why it matters
FinanceA 33x valuation step from a $105 million seed in sixteen months, on 30 to 40 robots and no named customer. The data-collection method is the asset being priced.
ExecutivesGlobalFoundries' sequence, four years of Spot then a gated Atlas evaluation with no unit commitment, is the procurement template. Earn the humanoid pilot with a simpler robot first.
ML engineersGenesis captures hand, glove and robot data 1:1:1 and says the hands wear out in weeks. That wear is a data-collection cost that no paper reports; budget for it before comparing data strategies.

Genesis AI announced a $500 million Series A on Tuesday at a $3.5 billion post-money valuation, co-led by Premji Invest, Khosla Ventures and Eclipse. Durable Capital, T. Rowe Price, Bpifrance, HSG, Nvidia, AMD Ventures, Zoom, Samsung Next and Ericsson participated, along with Eric Schmidt, Eric Yuan, Jerry Yang, Xavier Niel and Zach Kirkhorn as individuals. The company raised a $105 million seed in July 2025 from Eclipse and Khosla, so the valuation has stepped up roughly 33 times in sixteen months. Founders Zhou Xian, the chief executive, and Théophile Gervet, the president and formerly of Mistral, met at Carnegie Mellon; the company is split between Paris and San Carlos.

What Genesis has built

The company is vertically integrated in a way few robot-brain startups are. It trains a manipulation foundation model called GENE, runs its own simulator, Genesis World, builds its own dexterous hand, and makes a general-purpose robot called Eno that is not a humanoid in the classic sense. The distinctive piece is a sensor glove costing about $300 that more than 100 human workers wear while doing ordinary jobs, producing what the company calls 1:1:1 data: the same motion captured on the human hand, on the glove and on the robot's hand. The pitch is that this beats teleoperation, which is slow and expensive, and simulation alone, which has never closed the gap on contact-rich tasks.

TechCrunch's reporter, who visited the company, counted 30 to 40 Enos built, was told the production design is finalised and a US contract manufacturer will start next year, and saw a demonstration in which a robot dropped a box and struggled with a label. Xian said the hands last only weeks before they need replacing. Customers are a handful of undisclosed pilots; the rollout order is manufacturing, logistics and labs first, then hospitality and healthcare, then consumers.

What GlobalFoundries is doing instead

On the same day, GlobalFoundries said it will be the first non-automotive company, and the first chipmaker, in Boston Dynamics' Atlas early-adopter programme, with humanoid testing in its fabs beginning in early 2027. The first phase evaluates use cases, integration and facility readiness; task-specific training follows. No tasks or robot counts were disclosed. What was disclosed is the track record that earned the pilot: Spot has run at GlobalFoundries' sites in Malta, Burlington, Dresden and Singapore for more than four years, two Spots in Singapore have completed over 90,000 sub-fab inspections since April at a 99% mission success rate, and 875 routine inspections were redeployed to the robots in two weeks with a plan to double that.

Two prices for the same floor

Genesis is being valued on a data strategy before it has customers; GlobalFoundries is paying nothing for a humanoid until a quadruped has proven the floor. Both are rational. A model company's asset is its data flywheel, and Khosla and Premji are paying for the possibility that glove data is the one that scales. A manufacturer's asset is uptime, and a 99% mission rate on a robot that already works is worth more to it than any demo. The Unitree founder told the APEC CEO summit on Tuesday that the technical breakthrough in general robot models may be near but that production of millions of units is the real constraint; Genesis has not yet built fifty.

What to watch

Whether Genesis names a customer or an order before its next round; $3.5 billion on pilots is a high bar. Hand durability, because a replacement cycle measured in weeks sets the real cost of its data strategy. The first Atlas task list at GlobalFoundries, which will show what a humanoid is actually trusted to do in a cleanroom. And whether other Spot customers, which include most large utilities and several fabs, follow the same gated path into the humanoid programme.

roboticsfundingresearchNvidia
Written in our own words from the linked sources · updated 2026-11-18 · corrections: corrections@aibriefing.io

Get the brief in your inbox

One email every morning: the stories that mattered in AI, each with a one-line "why it matters" for your role. Free, no ads, unsubscribe any time.

Published daily by AI Briefing. We never share your address.
Like this? Get tomorrow's brief by email, free.