Humanoid Robots Are Coming To Factories. But Not The Way You Think

We’ve all seen the billions investors are pouring into humanoid robots. I’m currently tracking about $300 billion in ecosystem spending, including NEURA Robotics, which just raised $1.4 billion, Figure, Apptronik, and a brand-new European unicorn – billion-dollar startup – that I’ll be writing exclusively about tomorrow morning. Most of these robots, and others from top Chinese manufacturers like Agibot, UBtech, and Unitree are already working in factories, generally in demonstration projects, but not always. Agibot’s G2 is working on a live production line, something the company recently live-streamed.

We’ve even seen the first humanoid robot strike already: a Hyundai strike at least partially motivated by the fear of humanoid robots taking human jobs.

So here’s the big question: are we looking at the future of manufacturing, or is this all just another flash-in-the-pan tech hype cycle?

I put that to Jeff Burnstein, who has spent more than four decades at the Association for Advancing Automation and the last 19 as its president. He has watched the industry go from the first industrial robot arms to the AI-driven machines being built today.

He’s neither breathless nor totally dismissive.

"I think we’re heavy in the hype cycle," Burnstein told me. “We haven’t proven anything other than some production … the only humanoid in production in the U.S. that I’m aware of is an Agility robot in a GXO Logistics plant.”

But he’s also aware of the astonishing pace of innovation. In just a few years, we’ve gone from robots with claws to robots that can fry eggs and tape wires together.

“AI is changing this very rapidly,” he adds. “I understand that. The intelligence of these robots could be much better much sooner than I think. The dexterity could be much better much sooner.”

Most of what’s going on with humanoid robots in factories are pilot projects. And the pilots, Burnstein says, are landing at maybe 20% to 50% effectiveness. "That's not enough. Customers need 99-point-whatever in order to be certain of using these technologies."

(Interestingly, the UK startup Humanoid is currently achieving roughly 80% of human speed in some manual tasks, and is targeting 100% and higher.)

The point is that manufacturers don’t actually care whether a robot has two legs and a face. Or wheels, for that matter.

"What I hear from customers is, 'Look, we need a solution. If this is a better solution than existing robots or no robot, we’re good,'" Burnstein said. "We just need things that work, things that are proven reliable, affordable."

The single biggest gate on adoption? It’s not intelligence or dexterity, though both are needed. It's safety.

There is no humanoid robot safety standard yet, and until there is, Burnstein doesn’t see fast growth. "I don’t think insurers are going to be happy. I don’t think OSHA will be happy." Most if not all working humanoid robots are kept away from people with gates or fences. And one early accident with a humanoid could freeze the whole market.

Traditional industrial robots — massive, quick, purpose-built, accurate to many nines — already do enormous amounts of work, and that will keep growing. Humanoids will show up in factories, warehouses, logistics, farms, and homes. But we probably won’t see a factory with 10,000 humanoids, Burnstein says. There will be some; the question is how many.

Which brings up the jobs question, of course.

The narrative I’ve been using on my Next with John Koetsier podcast is that AI seems to be coming for the white-collar jobs and robots are coming for the blue-collar ones. And maybe we all end up on some kind of universal basic income.

Burnstein has been answering some version of the robots-and-jobs question since 1983, and his data cuts the other way.

A3 has tracked robot sales against U.S. employment for more than three decades. The pattern: when robot sales go up, unemployment goes down. When robot sales fall, unemployment rises. The logic is straightforward: companies that automate become more productive, win more business, hire more people.

The real job killer, Burnstein says, isn't the robot.

He frames the choice companies faced in the 1980s in three words: automate, emigrate, or evaporate. Many of them chased cheap labor overseas and evaporated anyway. Meanwhile there are nearly 500,000 unfilled manufacturing jobs today, a number projected to hit 1.9 million by 2033, according to Burnstein.

The problem isn’t a shortage of work. It’s that fewer people want to walk 10 miles a day hauling boxes in a warehouse. The jobs shift toward overseeing, installing, and maintaining the machines, in the same way that jobs such as "SEO specialist" or "iPhone app developer," didn't exist a generation ago.

If there's an urgent thread here, it's national strategy.

China holds roughly 54% of the world's industrial robots — a fleet of about two million — and it got there on purpose, through decades of five-year plans aimed at being both the largest user and the largest supplier of robotics.

The U.S. has been here before, and lost. Robotics was invented in America: the first robot was installed by Unimation in 1961, and 100-plus U.S. companies competed in the 1960s and '70s. Then Japan built a national strategy, Washington did nothing, and the industry migrated. America was left with essentially one robotics company.

Burnstein’s argument is that new form factors like humanoids, autonomous mobile robots and collaborative arms crack the field wide open again, and this is a second chance the U.S. can’t fumble. Encouragingly, legislation for a National Robotics Commission is being introduced, he says. Whether it goes anywhere is the open question.

Thirty years after the great debate between robotics pioneer Joe Engelberger and iRobot's Colin Angle, exactly one home robot has succeeded at scale: the vacuum cleaner. Angle bet on cheap and single-purpose, and he was right.

A $20,000 humanoid that folds your laundry is a tougher sell. "I can fold laundry," Burnstein said. “I don't need that.”