Amazon Bets on Robots: More Than a Million Machines and a Plan to Avoid Hiring 600,000 People

Amazon is no longer scaling through hiring alone. It is the second-largest corporate employer in the US, with about 1.5 million employees, behind only Walmart. In July 2025, the company reported a millionth robot in its network, and according to the New York Times, internal plans call for automating up to 75% of operations and avoiding the hiring of roughly 600,000 people by 2033. Amazon calls those documents incomplete and says they reflect the view of a single group. The direction is still clear: the company is betting on machines.

From carts to manipulators

It started with the acquisition of Kiva Systems in 2012 for roughly $775 million. Orange platforms carried shelving units to employees, who picked items by hand. That part of the job was long considered too hard to automate: items come in different shapes, and packaging is crumpled or shiny.

Now Amazon is going after exactly that part. The autonomous robot Proteus drives among people without fencing. The Sequoia system, according to the company, speeds up receiving and stowing of goods by up to 75% and cuts order processing time by up to 25%. Sparrow and Robin sort and move parcels. Vulcan, shown in May 2025, is equipped with tactile sensors and, according to Amazon, handles about 75% of the items stored in shelf bins. This is work that, until now, only human hands performed.

AI instead of a dispatcher

A fleet of a million machines needs coordination, so Amazon has invested in software too. DeepFleet is a generative model that manages how robots move around a warehouse. The company says it improves travel efficiency by about 10%. At a single site, that is minor. Across hundreds of warehouses, the gain adds up to millions of hours saved and hundreds of millions of dollars.

The company later showed Blue Jay, a system with several manipulators working together, and Project Eluna, an AI agent for operations teams. Public metrics for both are still scarce, so they are better judged by real-world deployment than by presentations.

The economics

The main argument for robots runs through the cost per item. According to the New York Times, Amazon expects to cut costs by about 30 cents on every parcel processed. Over 2025–2027, that comes to about $12.6 billion. At the same time, the company wants to double sales volume by 2033 without growing its workforce proportionally.

Key figures:

• Employees: about 1.5 million.
• Robots in the network: more than 1 million (as of July 2025).
• DeepFleet effect: about 10% in routing efficiency.
• Target share of automated operations: up to 75% (according to the NYT).
• Jobs not created by 2033: about 600,000 (same source).
• Expected savings: about 30 cents per parcel.

The logic is simple. Machines absorb the growth in orders, while people stay on tasks where automation is still more expensive or less reliable: non-standard items, robot maintenance, exceptions.

Practical takeaways

For logistics operators, the sequence is instructive: first automate movement, then grasping and sorting, and build an AI fleet-management layer on top. It is worth planning that layer early, because without it the gains from hardware quickly hit a ceiling.

For warehouse workers, the takeaway is less welcome. There will be fewer new positions, and requirements will shift toward equipment maintenance, working with data, and handling exceptions. Companies that retrain employees in-house will be more resilient.

For robotics startups, the window is open where Amazon has not yet built its own solution: flexible grasping, mobile humanoids, and software for mixed zones where people and machines work together. For regulators, the question is how to measure the impact: by the number of people laid off or by the number of jobs never created. The latter do not show up in any reporting.