Welcome, sci-fi realists…
This week, we would like to inform you all that a school district in Western New York is brave enough to attempt having an assistant humanoid teacher help out in one of its classrooms this coming Fall. The Salamanca City Central School District made the purchase from Realbotix, and is prepared to have the model (Sally) lend a hand.
The school made clear it won’t replace the actual teacher, but instead help tutor students and provide support when they need it. We can’t know how this experiment will turn out until their semester starts, and it’s too early to be cynical about the ways this could go wrong.
Also, picture your delivery driver pulling up, opening the van, and sending a yellow robot dog toward your porch with a package on its back. Boston Dynamics is testing Spot as a last-step delivery assistant, handling the awkward stretch from curb to front door that it calls the “porch gap.”
Spot carries parcels on a small conveyor attachment, walks them up to the house, then gently slides them onto the doorstep. Its legs could help where wheeled bots struggle: stairs, curbs, gravel, and the occasional suburban obstacle course disguised as landscaping.
Anyway, here's what you need to know about robotics right now:
Fujitsu partners with some of the biggest names in robotics to bring physical AI to multiple industries.
Agility Robotics builds a new Fremont, California facility to expedite physical AI production.
China dominates global humanoid development in surpassing 400 models.
Blackstone invests in South Korean provider Futuronic.
NVIDIA Brings Physical AI Down to Earth
NVIDIA arrived at SIGGRAPH with a new way to put more robot brainpower closer to the robot. Its Cosmos 3 Edge is a four-billion-parameter “world model” designed to run locally, helping machines understand a scene, predict what happens next, and choose an action without phoning a distant server first.
That matters in factories, warehouses, and anywhere a delayed robot reaction would be a very expensive interpretive dance. NVIDIA says developers can adapt the model for specific robots, sensors, and environments, then deploy it on Jetson and RTX hardware.
The flashiest demo was MotionBricks, a research system trained on more than 350,000 motion clips that drove both an animated character and a Unitree G1 humanoid. Cool? Absolutely. A ready-to-hire robot workforce? Not quite.
What this likely means for the future: More physical AI will run on-site, where fast reactions and privacy matter.
Our FYI read: The real story is less “robots got smarter” and more “robots may not need the cloud for every thought.” Watch the full keynote here.
Fujitsu Wants to Be Japan’s Robotics Switchboard
Fujitsu is already a big name in multinational IT tech alongside Dell or Samsung, but now it has big plans for physical AI expansion, too. Two of the “Big Four” robotics companies, Yaskawa and FANUC, are part of this new partnership, giving Fujitsu all the resources it needs to cater to the sectors it’s targeting.
Specifically, the partnership’s goals include competing in the new robotics “arms race”, addressing labor shortages in healthcare and manufacturing with AI-powered bots, and stressing less about aging workforces that will retire soon.
Fujitsu’s eventual “collaborative control” software platform will integrate NVIDIA’s physical AI with robot hardware from FANUC, Yaskawa, and Kawasaki Heavy Industries. This will create one big, powerful solution to meet those goals.
Readers of this newsletter can already guess that Fujitsu will apply physical AI to guide bots in manufacturing and warehouses, but the press release also mentions lesser-known use cases for healthcare. Here, bots could automate transporting pharmaceuticals and medical specimens, plus take on patient queries.
What this likely means for the future: Japan is assembling the pieces for more adaptable industrial automation, starting at home before aiming globally. For now, this is a set of business discussions and a proposed platform, not a fleet rollout. But it combines companies with serious robot-control experience and a clear reason to automate: Japan’s labor shortages and aging workforce.
Our FYI read: Because of how big these company names are and the unlimited resources they have to make this partnership a reality, we fully expect it to succeed in making Japan a key player in robotics.
🕹️ Robots in Action
A recent event that didn’t get as much fanfare as Automate 2026 in Chicago still showcased pragmatic robotics products from the Eastern Hemisphere. The World AI Conference (WAIC) in Shanghai featured some unique humanoid designs that will soon be ready for industry use.
Just one of many examples on display was a Youibot Robotics humanoid. We talked about AGIBOT’s G2 model before, and this model from Youibot has the same build: a humanoid torso that looks like it’s riding around on a rolling autonomous mobile robot (AMR), much like the ones warehouses use to move boxes.
The video below shows these models literally rolling up to crates, filling them up with random objects, and turning a full 180 degrees to bring them back. Check it out:
📰 Around the Horn
Agility Robotics opened a 60,000-square-foot Fremont facility where its teams will train, test, and develop new workplace skills for the company’s Digit humanoid.
Gritt emerged from stealth with $32 million in total funding and robots that helped eight-person crews install up to 4,000 solar panels per day, roughly five times the usual pace.
China leads global humanoid robotics development with more than 400 new models, more than half the global total, while its quadruped robots captured nearly 70% of worldwide sales during the first half of 2026.
Blackstone invests in South Korean robotics supplier Futuronic, backing the company’s expansion of the high-precision actuators and motion-control systems used in vehicles, humanoids, and industrial robots.
AGIBOT X2 series showcases embodied AI at the World Cup, demonstrating movement, interaction, and soccer skills during the global sporting event.
NVIDIA and Hugging Face added the GR00T 1.7 robot model, Isaac Teleop tools, datasets, and development workflows to LeRobot, giving researchers more open resources for training humanoids.
Ukraine’s Brave1 defense initiative launched a grant competition for combat humanoids, though wheeled and tracked robots remained the battlefield workhorses after completing more than 66,000 missions this year.
🚨 Bots Behaving Badly:
Now, what isn’t working so well right now?
As fascinating as it is to have a perfectly bloodless martial arts tournament with humanoid fighters, no one should expect their mechanical parts to stay intact. Case in point: the recent Ultimate Robot Knock-out Legend tournament in Shenzhen, China didn’t go without one humanoid losing its head, literally.
As one of EngineAI’s T800 humanoid models landed a head-high kick on another (its opponent), the kick proved powerful enough to nearly knock the opposing robot’s head clean off. Whether that was intentional or not is besides the point, it still intensifies the public’s fears about killer robots in the future.
Also, whether or not this counts as bad bot behavior is relative. Many onlookers still found the tournament to be a fun spectacle, especially since no actual human heads were involved. We could easily file this in the they-don’t-know-their-own-strength category, or we can continue our collective panic about killer bots. What we can all agree on is this: humanoid models are getting more dextrous, agile, and powerful every month.
📡 Industry Signals
It’s no surprise China continues to dominate humanoid production, but the 400+ figure is a revelation to most people, including observers of this space.
Other countries, such as Japan, have big plans to compete in this robotics arms race.
No one should expect physical AI to be rolled out as quickly as generative AI, but its development is nonetheless being set in motion all over the world, and fast.
Hi! I’m Cameron Hashemi-Pour a technical writer and editor who covers AI, robotics, machine learning, Industry 4.0, cybersecurity, and emerging technology. Previously, I wrote for TechTarget/Informa TechTarget and contributed scripts for the “Eye on Tech” YouTube channel.
That’s all for this week!