What if robots cost more than people? A new study suggests that the low-end jobs we assume will be automated might be too expensive.
VCG via Getty Images
Maybe robots will cost more than people. And just maybe, human jobs will be safe for longer than we think. According to a new report by a construction software company, replacing a single nursing assistant with a humanoid robot runs about $375,100 a year, nearly nine times the $42,200 these workers actually earn. And robots that could replace construction laborers, who make just under $50,000 per year, would cost almost $300,000.
“For years, the assumption was that low-paid, low-skill jobs would be the first to go,” a spokesperson from Planera, the software company that did the report, told me via email. “But the data shows the opposite.”
For a decade, the automation story has had a pretty clear villain and a clear victim. The villain was the robot (or those making them). The victim was the low-wage worker doing repetitive, “unskilled” labor: cashiers, assemblers, care aides … the person we were told would be first against the wall when the machines arrived.
This study says we have it almost exactly backwards. The reason: good robots are not exactly cheap.
The report looks at 30 of the most common jobs in America. It combines federal wage data with actual up-to-date pricing from automation and robotics vendors, then calculates the all-in annual cost of swapping each worker for a machine, including buying and installing the technology, keeping it running and paying the humans you still need to supervise it.
Turns out that the most expensive job to automate isn’t necessarily some high-end, high-tech profession. Rather, it’s a nursing assistant.
Here’s the data:
Chart of key professions and cost to automate them with robots. Data from Planera; chart by Claude.
John Koetsier
“The workers earning the least tend to be the ones doing the most physically demanding, human-facing work,” the Planera representative told me. “And that turns out to be exactly what machines struggle with most. It is white-collar roles that are now more exposed. Software developers, once considered untouchable, are already feeling it. Major tech companies have been cutting engineering headcount precisely because AI coding tools are getting good enough.”
Plot twist: humanoid robots are getting cheaper
Today all of the above is largely true. The plot twist, of course, is that technology and innovation don’t stand still. As of this moment, humanoid robots are the very worst they will ever be, and every single day, that statement becomes true all over again.
We see amazing hands, incredible humanoids, plus better and better AI every single week.
Just as importantly, not only is the technology getting better and better on both the hardware and software sides, we’re getting much better at reducing cost. Boston Dynamics’ newest Atlas is almost an order of magnitude simpler than its predecessor. 1X has so verticalized the supply chain for its Neo humanoid robot for the home that Dar Sleeper, head of design and product, told me the company’s costs are incredibly low.
And we haven’t even talked about Chinese manufacturing at scale. Baseline costs continue to shrink as the industry starts a long process of maturation and consolidation.
So there’s a lot of probably-justified hope from investors and entrepreneurs that eventually, robots will be cost-effective at scale. That’s why, according to my data, we’ve seen almost $300 billion in investment in humanoid robots so far. All of those investors believe they’ll get a return on their money.
But there’s no guarantees in life or business.
As Saxena, who has invested in humanoid robot companies, says, “the versatility, the dexterity of the human hand, that’s still unmatched by humanoids.”
The most likely result is that we will get humanoids who will be useful for some jobs, but that we’ll still need humans for many others, including the job of managing the robots, maintaining the robots, and filling the gaps where the robots fail.
Which probably means that the 10 jobs that are safe from automation are safe right now, but likely not forever. At least in their exact current form.