Category: Humanoids

SAVANNAH, Ga. — Boston Dynamics has opened the Robotics Metaplant Application Center, a permanent Atlas training facility inside Hyundai Motor Group’s Metaplant America campus in Georgia. The company announced the opening on 21 September, describing the site as the proving ground for a deployment that Hyundai has said will eventually put 25,000 Atlas humanoids to work across Hyundai Motor and Kia plants worldwide.

Why it matters: a factory that manufactures robot behaviour

RMAC is not a production line. Boston Dynamics calls it a “Robot Behavior Factory” — real automotive parts, real fixtures and real workcells, staffed by operators whose job is to generate training data rather than vehicles. Atlas learns there by three routes: teleoperation, reinforcement learning in simulation, and handheld Universal Manipulation Interface devices that capture human demonstrations without a robot in the loop.

The first tasks are deliberately unglamorous. Atlas is being trained on parts logistics and sequencing — picking components and placing them in assembly order — work that demands bimanual handling and awkward grasps but does not touch the moving line. Boston Dynamics has said component assembly itself is a 2030 target, and that it will begin exploring aerospace, semiconductor, logistics, food and beverage and life-sciences use cases in 2027.

The numbers behind it are the real signal. A second RMAC building launching next year will expand capacity roughly tenfold. A planned U.S. manufacturing facility is sized to build up to 30,000 robots a year. Boston Dynamics is relocating staff from its Waltham, Massachusetts headquarters to Savannah and hiring locally, while separately planning a $100 million Waltham expansion that the company projects will add 1,250 jobs by 2033. Taken together, that is a company building the supply chain for a humanoid fleet before the fleet has a proven job.

“Together with Hyundai Motor Group, we are pioneering physical AI at enterprise scale… RMAC reinforces our shared ambition to create a future where robots amplify human potential by making work safer and more productive.”

Zack Jackowski, Chief Product & Technology Officer, Boston Dynamics

The open question is throughput. Every humanoid programme now claims a data flywheel; almost none has published how many hours of collected demonstration translate into one reliably autonomous task. RMAC is Boston Dynamics’ answer to that question, and it will be judged on how quickly Atlas moves from sequencing bins to touching the line.

Sources

• Boston Dynamics — Boston Dynamics Opens Robotics Metaplant Application Center
• The Robot Report — Boston Dynamics opens Metaplant Application Center to train Atlas humanoid robots
• Axios Boston — Boston Dynamics opens Hyundai testing ground for humanoid robots

SALEM, Ore. — Agility Robotics has unveiled Digit 5, the fifth generation of its industrial humanoid, with a 40% payload increase to 50 lb (22.7 kg), reach extended to 7.2 feet, and a 90-minute battery that recharges in 9 minutes. That 10:1 run-to-charge ratio is the headline engineering claim: Agility says it yields more than 20 hours of work per 24-hour day.

The order book and the income statement

Agility says it holds more than $300 million in multi-year customer orders as of May 2026, including one commitment for 1,000 robots over three years. Against that, the company reported $1.78 million in net sales for 2025 and an operating loss of $140.2 million. The gap between a signed order book and recognised revenue is the central number in humanoid robotics right now, and Digit 5 is the machine meant to close it. Early access begins in the first half of 2027, with general availability by the end of 2027 across North America, the EU and the UK.

The hardware changes are aimed squarely at that transition. Digit 4’s backwards “ostrich” legs are gone, replaced with human-like knees and proprietary cycloidal actuators built for heavy lifting. The robot weighs 284 lb and stands 5 feet 11 inches. End-effectors are now swappable on ISO-standard flanges.

The more consequential change is safety. Agility says Digit 5 is the first humanoid to pass an independent OSHA field evaluation, and the company contributed to ANSI/A3 TR R15.108 and ISO 25785-1, the first international safety standard for humanoids. A separate safety controller and 360-degree human detection let the robot steer around people, stop, or sit down, with visual and audio cues signalling intent. In practice that means working without a cage — which is what actually determines whether a humanoid can be dropped into an existing facility or requires rebuilding one around it.

“Digit 5 is removing a major barrier to scaling humanoid robots in industrial environments.”

Peggy Johnson, CEO, Agility Robotics

Manufacturing runs through RoboFab in Salem, Oregon, which Agility has said can build up to 10,000 robots a year. That capacity figure, the order book and the 20-hour uptime claim are all company-stated and not independently verified. The OSHA evaluation and the standards work are the parts with outside parties attached.

Sources

• Agility Robotics — Agility Unveils Digit 5 Humanoid Robot Built for Cooperatively Safe Work at Scale
• GeekWire — Agility’s new Digit 5 robot lifts 50 pounds, works 20+ hours a day
• The Robot Report — Agility’s Digit 5 humanoid has new legs, batteries, safety upgrades

ANALYSIS — Every few years someone declares that the home robot has finally arrived. Usually they are wrong. But the ground is genuinely shifting: cheaper sensors, capable AI, and real manufacturing scale are converging at the same time. Here is what the next decade of home robotics will actually deliver, measured against what the industry is currently spending and shipping.

The money is now real, and it is going into brains

The clearest signal is not a product launch — it is a compute bill. Figure has committed to up to 100,000 Nvidia GPUs in a $3.5 billion deal with Nscale, with stated intent to scale past $6 billion, all of it to train Helix, the model that drives its humanoids. Nobody spends that on actuators. They spend it on perception and reasoning.

That is the whole story of this decade in one line item. The mechanical parts of robotics have existed since the 1980s. What was missing was a brain flexible enough to handle a cluttered, unpredictable room. The body was never the bottleneck.

From single-task gadgets to general manipulation

The robots already in homes — vacuums, mowers, pool cleaners — each do exactly one thing. The frontier is general-purpose manipulation: picking up, carrying and handling everyday objects the machine has never seen before. That capability is what turns a gadget into a helper, and it is the single hardest problem in the field.

Manufacturing scale arrived first in China

Capability means nothing without the ability to build. The second World Humanoid Robot Games drew 666 teams from 16 countries fielding 2,056 humanoids — triple the prior year. Days earlier, the World Robot Conference opened in Beijing with 150 product debuts. And the public markets have noticed: Unitree’s Shanghai listing rose nearly sixfold on debut.

What will not happen yet

Honesty matters in a field this hyped. A robot that folds laundry, cooks dinner and minds the kids is not arriving this decade. Even the most aggressive commercial timeline on record is modest: Tesla has told analysts that Optimus sales could start in late 2027, with the Gen 3 design finalized and a line going into Fremont. That is a starting gun, not a finish line.

Progress will look incremental and unglamorous. A robot that reliably does three useful things beats a prototype that impressively fails at thirty. The winners will earn trust one dependable task at a time.

Why this decade is different

Three curves are crossing at once: the cost of capable hardware is falling, AI capability is rising fast, and the manufacturing base to build at volume finally exists. When those align, adoption stops being linear. The companies laying groundwork now — in perception, safety and real-world reliability — are the ones that will define the decade.

Zip Robotic covers robotics, automation and the companies building them. Read the latest.

FREMONT, Calif. — At a JPMorgan-hosted tour of its Fremont factory, Tesla told analysts that the design of Optimus Gen 3 is finalized, its supply chain is “essentially locked in,” and a production line is being installed in the facility that until earlier this year built the Model S and Model X.

The headline: external commercial sales of Optimus could begin “as early as” the second half of 2027. Tesla repeated its long-term capacity goals of 1 million units a year at Fremont and 10 million in Texas.

The context

Electrek, which covered the tour, notes that the Gen 3 reveal has slipped repeatedly and that Elon Musk has acknowledged no Optimus units are yet doing useful work inside Tesla. The Model S/X line was torn down in April to make room — a real, expensive commitment — but “as early as” is doing a lot of work in that sentence.

For comparison: Unitree shipped more than 5,500 humanoids in 2025 and Figure has deployed units with paying customers. Tesla’s bet is that when it does ship, it ships at automotive scale. That has never been done in robotics.

Sources

• Electrek — Tesla JPMorgan Fremont tour: Optimus 2027
• Electrek — Model S/X line converted for Optimus