On Amazon's Robotics Dreams

A recent New York Times piece sounds the alarm about Amazon automating away work, but the conclusion is actually not that dramatic

I have a longer piece coming out soon about Amazon’s robotics game and what it means for labor in the coming years. But I thought a brief response to a recent New York Times piece, with the alarming title “Amazon Plans to Replace More Than Half a Million Jobs With Robots,” deserved a quick response.

The big reveal is in the title, but some of the details demand more investigation. These include:

• that “Amazon’s automation team expects the company can avoid hiring more than 160,000 people in the United States it would otherwise need by 2027. That would save about 30 cents on each item that Amazon picks, packs and delivers to customers.”

• that the company plans “to avoid adding to its U.S. work force in the coming years, even though they expect to sell twice as many products by 2033. That would translate to more than 600,000 people whom Amazon didn’t need to hire.”

• “that Amazon’s robotics team has an ultimate goal to automate 75 percent of its operations.”

• that ”Amazon plans to copy the Shreveport design in about 40 facilities by the end of 2027…. That facility currently has roughly 4,000 workers. But once the robotic systems are installed, it is projected to process 10 percent more items but need as many as 1,200 fewer employees, according to an internal analysis.”

Now, the company says that this is all according to one team within Amazon and that it doesn’t reflect the positions of the company as a whole, and I think this is actually a fair rebuttal, as far as it goes, given how famously decentralized Amazon is. I would love to be able to see the internal documents on which this article is based for myself, as I’m guessing that there is some nuance to some of the claims here that would be interesting to tease out. For instance, both the Sequoia and Vulcan systems can supposedly handle about 75% of baseline pick/stow work (not totally sure about that for Vulcan with the pick numbers, but certainly stow). Maybe this is where that 75% number comes from, but that is a very different claim from “automating 75% of its operations,” which is too vague to make sense of.

The key claim is about doubling sales while avoiding hiring 600,000 people. Again, I would need to see the actual documents to know for sure here, but on the face of it, that means in practice that Amazon’s American workforce will have 600,000 more people in it by 2033. Right now that workforce is made up of 1.2 million workers. Doubling sales would, at current productivity levels, involve doubling the workforce size to 2.4 million. This is not (just) a software company: if they want to deliver twice as many packages in their current arrangement, they need twice as many people to move stuff (though they can always gigify work, which they’re attempting to do as well, but this is a different issue than automation). With a larger penetration of the latest generation of fulfillment tech, that number is lowered by 600,000, in their estimation, and so we’d be looking at a US workforce of 1.8 million in 2033.

Maybe you don’t like my baseline assumption that the workforce would otherwise be doubled: certainly corporate doesn’t need to grow at the same rate as warehouse, maybe there’s natural reorganizational efficiency that doubling package volume would bring even without automation upgrades. But I find it difficult to believe that at current technology deployment levels, doubling sales would not mean adding significantly more than 600,000 workers.

This all adds up to a much less shocking scenario than the article presents, and it’s actually a less dramatic displacement than what occurred over the pandemic. Amazon only reports its global workforce size, which dropped from 1.608 million employees in 2021 to 1.556 million employees in 2024. They don’t break out the precise size of their American workforce, but let’s just say it stayed stable around 1.2 million during that time. In that same period, AMZL’s US package volume jumped from 4.8 billion to 6.3 billion. If its US workforce has grown at the same size during this period, Amazon would have employed 1.575 million people in 2023. Thus, they displaced about 375,000 US workers in just three years.

Again, maybe you don’t want to count corporate and AWS workers: I’ve previously estimated the US warehouse workforce to be comprised of about 914,000 workers. If we just counted the growth of warehouse workers in line with package volume growth, Amazon would have been employing 1.2 million warehouse workers in 2024 - thus, a displacement of 286,000 warehouse workers. No matter how you slice it, Amazon replaced hundreds of thousands of jobs during the pandemic. (I’ll explain more why that happened, and why I don’t think they’re likely to repeat such dramatic displacement in the coming years, in that forthcoming longer article.)

The Shreveport bit actually makes me less worried than I was before about what those Gen 12 Fulfillment Centers mean for labor. The Shreveport facility until not that long ago only employed about 1,400 people, far less than the average 3,000+ of a normal ARS facility. If, once fully staffed up, this new generation is only employing at 70% of the levels of other AR facilities, that is no employment apocalypse, and given how much capex is involved in making them happen, I have a real difficult time believing that Gen 12 is going to be fully operational in 40 facilities by 2027 (though the One Big Beautiful Bill gave them a lot of money to do so).

My conclusion from all of this is that these internal documents a) represent only the thoughts of one team, b) are presented vaguely enough here that it’s difficult to specify what precisely is being claimed, and c) don’t seem to reveal much more than we already know about Amazon, which is that it is rapidly displacing work through its automation upgrades but in such a way that is largely masked by company growth.

It’s worth saying also that Amazon is probably not very concerned about correcting the narrative of this article, though it officially contradicts much of what they say about their “cobots,” because it exercises a clear disciplinary function on labor. “We are on the cusp of automating away your job” is a thought that they’re happy for every Amazon worker to have lingering in their mind somewhere. At the workplace level, sometimes it is indeed easy to replace baseline pickers and stowers, but not quite as easy as you’d think. Amazon spends a lot of time, money, and effort picking locations for their facilities where they are sure they can staff up in the ways they want. But at the macro level, they’re going to need a lot of people for the foreseeable future, and it’s good to keep this in mind amidst all the AI apocalypse talk.