AI & Tech Issue #3 ·

Robots at $5 an Hour: What Will Humans Do?

The question is shifting from how well robots perform to what humans do.

Oswarld (Kwangseob Ahn) · GTM & AI strategist working the seam between technology and business

Robots at $5 an Hour: What Will Humans Do?

Opening

Subscriber, let me start with one number: $5.71. That is the hourly wage of a robot now working around the clock in an American warehouse. It is the cost per hour after including energy and maintenance. In Korean won, it is about 7,800 won.

What is the average hourly wage for an American warehouse worker? According to PayScale, it ranges from $13.90 to $22.57. In states like California, the minimum wage alone is $16.90.

The age in which robots work for one-third or one-quarter the cost of humans has already arrived. But what I want to talk about today is not the number itself. It is the structural change this number produces — how employers’ calculations are changing, and how the grammar of the labor market is being rewritten.

Why Robots, and Why Now?

Robots working in factories is nothing new. But the change now under way is different in kind. Three things are moving at once.

First, production costs are falling sharply. One estimate puts the average price of an industrial robot at about $46,000 in 2010 and about $10,856 in 2025. In a decade, the price has fallen to roughly one-quarter of its former level. Small autonomous mobile robots, or AMRs,1 can be deployed for around $10,000.

Second, robots have acquired “eyes” and “brains.” Earlier robots were simple machines that followed fixed routes. Thanks to advances in multimodal AI2 and computer vision, today’s robots can perceive their surroundings, make judgments, and adapt. The fact that a small chipset such as Nvidia’s Jetson series can now handle simple multimodal tasks is a major shift. It has become possible to put a cheap “brain” on a robot.

Third, training costs are falling, too. In the past, training a robot required repeated trial and error in a real physical environment. Now, as virtual training through digital twins3 and simulators becomes commonplace, the time and cost between development and deployment have fallen dramatically.

When these three forces come together, something important happens. From an employer’s point of view, the ROI calculation changes completely. Robots work 24 hours a day. They do not get sick, do not strike, and do not require workers’ compensation insurance. If they break down? Replace a part. In Charlie Chaplin’s Modern Times, humans were depicted as components of a machine. Now actual machines are taking that place.

A Glimpse of the Future From Amazon

The company that shows this shift most clearly is Amazon.

As of mid-2025, Amazon had deployed more than 1 million robots across its logistics facilities. That is almost the same number as its roughly 1.56 million human employees worldwide. The robot fleet began with 1,000 units after Amazon acquired Kiva Systems in 2012. It reached 200,000 in 2019, 520,000 in 2022, and has now passed 1 million.

According to internal documents obtained by The New York Times, Amazon’s robotics team aims to replace more than 600,000 jobs with robots and automate 75% of overall operations. At Amazon’s automation pilot facility in Shreveport, Louisiana, the workforce is said to have fallen by about 25% after robots were introduced.

What is interesting is Amazon’s language. In internal documents, the company deliberately avoids the word “automation.” It does not use the term “AI” either. Instead, it calls the technology “Advanced Technology,” and refers to robots as “cobots,” or collaborative robots. This is framing4 designed with social resistance in mind.

At the same time, Amazon is offering an alternative. It is expanding the role of “robot technicians,” who manage and maintain robots, and the hourly wage for this role is $24.45, higher than that of general warehouse jobs, at about $19.50. The company also runs training programs so existing employees can transition into these roles.

But this is where we need to pause. According to The New York Times, at one Amazon warehouse densely populated with robots, only about 100 of the 2,500 workers were responsible for robot-management tasks. It is true that new technical roles are being created. Whether they are being created at a scale large enough to offset the jobs disappearing is a separate question.

Why Not Humanoids?

Let me clear up one common misunderstanding. When many people hear “robot,” they picture a humanoid shaped like a person. But the robots actually doing the heavy lifting in industry look nothing like that.

The most efficient industrial robots are wheeled autonomous mobile robots or multi-jointed robotic arms. Some companies, such as Figure AI, even operate robots with 16 arms. The same logic applies to quadruped robots and other special-purpose machines.

The reason humanoids are inefficient is straightforward. A human-like form may feel familiar, but in productivity terms it is hard to beat a robot designed for a specific task. Battery life is another serious problem — at best 6 to 7 hours, and sometimes only about 2 hours. In industrial settings, robots that remain connected to AC power are far more practical.

The rental cost of Agility Robotics’ humanoid robot Digit is also telling: $30 per hour. At that price, it costs about the same as hiring a human worker, or even more. Humanoids are likely to shine only in areas where interaction with humans matters, such as household services or the senior-care industry.

Oz’s Lens

To be honest, I do not want to say this change is frightening or that it must be stopped.

In building GTM strategies, I have seen this pattern countless times. Whenever a new technology appears, fear comes first: “Jobs will disappear.” Then optimism follows: “New jobs will be created.” Reality has always been somewhere in between.

This time, however, one thing is different: speed.

Past waves of automation unfolded gradually over decades. Society had time to adjust. Now AI and robotics are advancing at the same time, and rapidly. AI-generated songs are being streamed millions of times on Spotify. AI-written text is flooding the internet. Robots’ hourly costs are falling below the human minimum wage. All of this has happened within just 2 or 3 years.

In Sweden earlier this year, an AI-generated song topped the Spotify chart and was streamed more than 5 million times. This was not the result of algorithmic recommendation alone; people actively searched for it and listened. Breaking Rust, a fictional AI-generated band, has recorded 2.5 million monthly listeners. The age of “AI slop” that began with writing is now expanding into music, video, and even physical labor.

I think of this as a wave. People who can ride the wave become surfers. People who merely watch from the beach get swept away when the tsunami comes. People who float like buoys will drift all the way into the middle of the Pacific.

The core question is this. Not “Will robots take my job?” but “What can I do that robots cannot?” We need to ask it now.

Closing

Here is the summary.

The hourly cost of robots has fallen below one-third of the human minimum wage. Amazon is already operating 1 million robots and planning to automate 600,000 jobs. This is not the distant future. It is happening now.

Two hundred years ago, during the Luddite movement, textile workers smashed machines. What we need to do now is not smash the machines, but redefine our value in areas machines cannot replace.

In the next newsletter, I will look at the other side of this structure — the uniquely human capabilities that robots and AI cannot replace, and how we can develop them.

📎 References & Further Reading

The author, Kwangseob Ahn, is a professor in the Department of Business Administration at Sejong University and lead consultant at OBF(Oswarld Boutique Consulting Firm). At the university, he teaches statistics and data analysis, including business data management and business analytics; in the field, he leads GTM strategy and AI strategy consulting, designing the intersection of technology and business. He has published an academic paper on memory architecture for AI dialogue systems (HEMA), and runs Daily Arxiv, a project that curates global AI papers every day. He completed a master’s program at Korea University’s Graduate School of Management of Technology and the KMBA. He is the author of 《People Who Outsource Thinking: Homo Brainless》.

Footnotes

  1. AMR (Autonomous Mobile Robot): A robot that perceives its surroundings and finds the optimal route on its own, without following a fixed path. Whereas a conventional AGV, or automated guided vehicle, follows lines on the floor, an AMR moves by making its own judgments. ↩

  2. Multimodal AI: AI that understands and processes multiple forms of data at once, such as text, images, and audio. Applied to robots, it allows them to understand what they see through cameras, listen to voice commands, and decide on appropriate actions. ↩

  3. Digital Twin: A virtual replica of a physical environment in the real world. Testing robots first in this virtual environment, instead of training them inside an actual factory, can greatly reduce cost and time. ↩

  4. Framing: The phenomenon in which people’s perception changes depending on the frame through which the same fact is presented. “Automation” evokes job losses; “advanced technology” feels more like innovation. ↩

Your take shapes the next issue

Reply with your experience or perspective — the best responses feed into future issues.