CES has regularly featured robots over the years, but we’ve never really seen anything pivotal. CES 2025 marked a change in this area. “The ChatGPT moment for robotics is just around the corner,” said Nvidia CEO Jensen Huang in his keynote, and we couldn’t agree more. And while attention was focused on LLMs, the field of industrial robotics has been unleashed like never before. According to World Robotics 2024, the International Federation of Robotics’ recent report, 4.3 million units were deployed in factories worldwide as of Q3 2024, a number that’s increasing at a clip of half a million units per year. This is double from 7 years ago, and the trend is accelerating.

These industrial robots don’t make it to CES, but the past 10 years’ extraordinary innovation around machine intelligence and 4D movement is starting to trickle down in significant ways, as we witnessed in Las Vegas.

These May Be the Droids We’re Looking For

Even the myriad of robot-ish companions felt a lot more “finished” than years past. In the “silly but cute” category, the hands-down winner was the clip-on bot Mirumi from Yukai Engineering (below). More accessory than robot, it’s in the crowdfunding stage right now, with a $70 retail price target, but we can easily see this becoming a must-have in Asia.

Samsung promised that its much-anticipated home companion Ballie, unveiled in 2020, will finally be commercialized in the first half of 2025. Meanwhile, like everyone else, we fell in love with TCL’s Ai Me (pronounced “Amy”), a companion robot concept that is by far the cutest and most clever take we’re ever seen in the “Pixar-inspired” genre.

Like Ballie, Ai Me (below) is a smart home control center. Unlike Ballie, it’s incredibly cute and cuddle-ready because the robot is separate from its egg-shaped “carrier.” which navigates its environment with impressive mobility, thanks to an integrated camera. One of these days one of these will hit it big and this one may be it.

A big unlock in the companionship segment is LLMs, a serious game-changer for conversational agents, so the level of experience developers can build into robots now is orders of magnitude greater than even last year.

One example was Realbotix’s somewhat strange but buzzy “female companion” robot Aria. It’s still pretty rough around the edges, especially for the $175,000 price tag, and the conversational elements still are not ChatGPT-grade, but we can clearly see how the systems will improve in the near future. Pretty niche play, but it’s going to get better fast.

Leap Forward in Mobility and Dexterity

More seriously, we’re keeping a close eye on a sub-category of robots that will likely revolutionize home management and home care over the next 5 years. In this category, the 20 years and billions of dollars invested by Boston Dynamics and others in 4D mobility is trickling down to the consumer level. This year we saw a pretty impressive jump in “affordable” real world navigation and mobility.

On the higher end, OpenDroids R2D3 “home helper” robot turned a lot of heads at CES. Labeled “Roomba on Crack,” this $60,000 robot is capable of handling such complex 4D tasks as folding laundry and washing dishes. The dexterity seems impressive, and we love the focus on a small handful of tasks, which in robotics directly translates into robustness.

In the same category, it was hard to miss the star of CES 2025: Roborock Saros Z70 (below), a Roomba clone with a mechanical arm capable of picking up debris in its way. Sure it’s a bit gimmicky, but it’s apparently the first mass-produced robotic arm, and presents a clear innovation that, for once, isn’t chasing a use case.

Much more technically sophisticated in our book was the Lymow One robotic lawnmower. With tank wheels and mulching blades, it seems very technically sound and robust, especially when it comes to navigating obstacles and traversing 45-degree slopes. And for $2,000 that’s not bad, especially since it can cover almost 2 acres per day.

We were also impressed with South Korea’s Tesollo robotic grippers, which represent both an order of magnitude greater capability for a fraction of the price of previous such systems. We can’t wait to see them in the systems of their clients (Samsung, LG, Hyundai).

But by far the most impressive “price-performance” mix was Unitree Robotics‘ $1,600 robot dog (below), a clone of Boston Dynamics Spot, which amazed us with its agility. This is the future of policing for sure.

Clearly China is taking a leading role in cheap and capable at-home robots, which will be certain to raise some serious national security eyebrows. But whether it’s China or another country, expect these to share your living space in the next 10 years. The tech is ready for prime time.

Nvidia Makes a Big Play with Cosmos

To scale, the whole field will need a lot more data. And as always, Nvidia is leading the way with its Cosmos AI platform, announced with fanfare at CES. Cosmos is a set of AI tools, which, together with Omniverse, aims to accelerate the development of physical AI systems, particularly for autonomous vehicles and robots.

From the early documentation, it looks like users can use Omniverse to generate simulations, which Cosmos will transform into photo-realistic video imagery aimed at training any agent system (like a robot or self-driving car) meant to navigate the world in 4D. “Take thousands of drives and turn into billions of miles.” is how Jensen Huang summed it up.

The platform features generative “world foundation models” (WFMs), advanced tokenizers, and an AI-accelerated data processing pipeline. Cosmos enables developers to generate synthetic training data and build custom models more efficiently, processing 20 million hours of videos in just 14 days using Nvidia’s Blackwell platform.

The platform’s new tokenizer offers 8x more compression and 12x faster processing than current solutions, while incorporating trustworthy AI principles with built-in guardrails to mitigate harmful content. Cosmos WFMs are available under an open model license, with early adopters including major companies like Uber, Toyota, XPENG, and several robotics firms.