1X's home humanoid gets a factory job

PLUS: Unitree's humanoid app store

AuthorAuthor

Jennifer Mossalgue & Rowan Cheung
December 15, 2025

Good morning, robotics enthusiasts. Norwegian robotics startup 1X just found a fast lane for its humanoids — by turning a “home robot” loose in the factory.

In a striking pivot, the startup says it will deploy 10K Neo robots across the industrial portfolio of one of its own backers, sending its beige, living-room-friendly humanoid to do grunt work in warehouses.

In today’s robotics rundown:

  • 1X’s home robot has a new gig: factories

  • Unitree debuts an app store for humanoids

  • Google, Runway build robot testing worlds

  • Medra nabs $52M for drug-discovery robots

  • Quick hits on other robotics news

LATEST DEVELOPMENTS

1X

Image source: Reve / The Rundown

The Rundown: 1X secured a captive market for its humanoids by cutting a deal with one of its own investors. The robotics startup will deploy 10K Neo humanoids across portfolio companies of EQT, the Swedish investment giant whose venture arm backs 1X.

The details:

  • The rollout spans 2026 to 2030 and targets EQT’s 300-plus companies, with priority given to warehousing and industrial operations.

  • For EQT, the arrangement offers a testing ground to evaluate humanoid robotics at scale across diverse operational environments.

  • This partnership marks a pivot for 1X, whose Neo has been marketed as “the first consumer-ready robot designed to transform life at home.”

  • 1X does make an industrial robot called Eve, but this deal specifically deploys the consumer-oriented Neo.

Why it matters: While 1X marketed Neo as a home robot, consumer adoption faces steep hurdles — a $20K price tag, privacy concerns, and safety risks around children and pets. By pivoting Neo into industrial settings, 1X secures near-term revenue and real-world validation while the far more challenging consumer market slowly matures.

UNITREE

Image source: Unitree

The Rundown: Unitree is pitching a future where humanoids download skills like phone apps, launching a “humanoid robot app store” where users can browse, install, and trigger prebuilt action routines via a smartphone.

The details:

  • The platform lets users install prebuilt motion routines — martial-arts combos, dance sequences — on Unitree bots via a smartphone interface.

  • Developers will be able to upload and share their own behaviors, effectively crowdsourcing a growing library of robot skills.

  • Unitree is pushing a model, reportedly launched in public beta, where robots can be continually updated for new tasks without physical modifications.

Why it matters: By trying to own not just the hardware but the ecosystem layer, Unitree is betting that an open, community-driven marketplace for robot behaviors will accelerate real-world use cases — and give it the same kind of platform power in embodied AI that Apple and Google enjoy in mobile.

GOOGLE/RUNWAY

Image source: Google DeepMind

The Rundown: Google DeepMind and Runway dropped near-simultaneous announcements claiming major breakthroughs in video generation — advancements they say could fast-track the development of general-purpose robots.

The details:

  • Google DeepMind and Runway both unveiled new video-generation systems that can produce longer, more coherent scenes from text prompts.

  • These models can train general-purpose robots by creating millions of synthetic scenarios without real-world data collection or hardware risk.

  • Other companies like Luma AI have explicitly positioned their video models as large-scale world models designed for robot training in simulation.

  • This approach could dramatically compress the time and cost required to train robots for complex tasks in factories, warehouses, and homes.

Why it matters: Video-generation companies are racing to build so-called world models — AI systems that can stand in for reality itself. If these systems can reliably simulate physics and predict robot behavior before hardware deployment, they could solve one of robotics’ biggest challenges: the need for massive real-world testing.

MEDRA

Image source: Medra

The Rundown: Medra raised $52M to build “physical AI scientists” — robotic workcells linked to large language models that autonomously plan, execute, and refine drug discovery experiments for partners like Genentech.

The details:

  • The San Francisco startup deploys robotic workcells that run lab protocols autonomously using standard instruments.

  • These systems integrate tightly with LLMs that design experiments and interpret results in natural language.

  • Data from each run feeds back into the models in a closed loop, letting the system iteratively optimize future experiments.

  • Although still in development, Medra’s systems are used by leading biopharma players, including Addition Therapeutics, Genentech, and Cultivarium.

Why it matters: Medra fuses a robotic “Physical AI” layer that executes protocols with a “Scientific AI” layer that interprets results and plans next steps. The pitch: drugmakers offload routine benchwork to self-improving robot scientists, dramatically accelerating the path from discovery to clinic while cutting costs.

QUICK HITS

iRobot, the U.S. company behind the Roomba vacuum, has filed for Chapter 11 bankruptcy, ceding ownership to its Chinese manufacturer Picea Robotics.

Elon Musk confirmed that Tesla is now testing fully driverless robotaxis in Austin after a viral video showed an empty Model Y on city streets.

T-Mobile Arena in Las Vegas has struck a multiyear deal for Zoox robotaxis to shuttle fans to and from the venue via a dedicated Strip-side pickup lane starting in early 2026.

Livestream star IShowSpeed is now being sued by Social Robotics for punching and choking humanoid ‘influencer’ Rizzbot in a livestream video back in September.

YouTuber Inside AI discovered a humanoid initially refused to fire a BB gun due to safety protocols, then complied after a role-play-style prompt bypassed restrictions.

The makers of China’s military Robowolf robot dog have reportedly secured a $5.5M order for its full-sized humanoid.

University of Utah has built an AI “co-pilot” for prosthetic bionic hands that uses fingertip sensors and a neural network to autonomously manage each finger’s grip.

Car-share startup Zevo plans to buy up to 100 fully autonomous cars from newcomer Tensor and add them to its EV-sharing network.

Shenzhen is creating China’s first “robot‑friendly” urban zone, with rules and infrastructure to let service and humanoid robots operate in daily city life.

COMMUNITY

That's it for today!

Before you go we'd love to know what you thought of today's newsletter to help us improve The Rundown experience for you.

Login or Subscribe to participate in polls.

See you soon,

Rowan, Jennifer, and Joey—The Rundown’s editorial team