The real way to tell if Amazon's AI spending is paying off

By Jurica Dujmovic

Investors focus on Amazon's cloud business, but the most important indicator might lie in the company's warehouses

Amazon has made big investments in automation.

Amazon's simplest case for artificial intelligence may be a cheaper trip to the customer's door.

The company (AMZN) sells computing power through Amazon Web Services, but it also uses AI inside a business with an enormous, recurring physical task: moving merchandise.

Every order that leaves a warehouse carries a cost. Better software can lower it by helping existing equipment handle more work, spreading the cost of buildings and machinery across more orders. If the software delivers, the savings should eventually show up in Amazon's margins, a measure of profitability.

The next test comes on Oct. 29, when Amazon reports third-quarter results. I will be looking for evidence that its warehouse technology is improving the economics of fulfillment.

Amazon has already given investors some numbers to test. It deployed its millionth robot in 2025, alongside the introduction of DeepFleet, an AI system designed to coordinate robot movements. The company said the system would improve the fleet's travel efficiency by 10%. Less waiting and fewer wasted trips could let the same machines move more orders during a shift.

It has also described a target at its advanced fulfillment center in Shreveport, La.: improving its "cost to serve" (the cost of getting merchandise to customers) by 25% during peak delivery seasons. The facility combines inventory storage, mobile robots, robotic arms and packaging technology.

These figures describe different opportunities: more efficient robot travel and lower costs at a particular facility. Neither establishes a companywide profit gain. The next step is for Amazon to show how widely it can reproduce the improvements, and what deployment costs.

The company's own rollout plan gives investors something to measure. On its July earnings call, Amazon said it expected to more than double its robotic-arm fleet in 2026. Thursday's report offers a chance to assess progress toward that year-end goal, including whether additional arms are increasing the volume that existing facilities can handle. Not every deployment lasts, though. In February, Amazon pulled Blue Jay, a multiarm system it had unveiled four months earlier, from its operations.

The share of orders benefiting from the technology matters more than a fleet count alone. Installing equipment at a busy facility could have a larger financial effect than adding the same equipment at a lightly used site. Investors should look for evidence that Amazon is putting the machines where they can earn the strongest returns.

Small improvements can matter at Amazon's scale. Its North American segment generated $116.18 billion in second-quarter sales and $9.12 billion in operating income, an operating margin of approximately 7.9%. Even an additional half-percentage-point margin on that revenue would represent roughly $581 million in quarterly operating profit, assuming sales stayed unchanged.

Measuring the contribution requires care. Amazon's North American segment includes advertising and subscription services, alongside retail. A stronger segment margin could reflect a more profitable revenue mix. Management's explanation of fulfillment costs matters as much as the headline margin.

Within fulfillment, AI also works to complement broader operational changes. In May, CEO Andy Jassy described placing inventory closer to customers and combining more purchases into each shipment. Both measures can lower costs by shortening shipping distances and reducing packaging, making it even harder to determine how much of the savings comes from newer AI systems.

A major potential source of savings is the labor required for each item. In October 2025, the New York Times reported internal estimates, as summarized by the Verge, that automation could let Amazon avoid more than 160,000 prospective U.S. hires by 2027, saving about 30 cents per item. Amazon said the documents reflected one team's perspective, not its overall hiring strategy.

Those are reported projections, not demonstrated savings. But they explain an important mechanism: Amazon could handle growth with fewer additional labor hours. Avoiding future hires differs from laying off existing employees.

The staffing mix would change, too. Amazon says advanced robotics facilities will require 30% more employees in reliability, maintenance and engineering roles. More specialists can coexist with lower overall labor requirements. The financial question is whether the reduction in routine work outweighs the added technical support and equipment costs.

Equipment must be purchased, maintained and eventually replaced. Retrofitting a working building can disrupt operations. Even a system that processes more orders per hour may disappoint financially if its installation costs too much or demand leaves it underused. Investors need the return on the whole installation.

The timing matters as well. Amazon may pay for equipment before it reaches normal operating capacity, while accounting spreads the equipment's cost over its useful life through depreciation. So the margin can look healthy while the cash spent up front has yet to be earned back.

Successful automation also gives Amazon choices about where the savings go. It could retain more profit, improve delivery speed or hold down prices. Faster delivery and lower prices can pay off if they bring customers back more often, but shareholders should look for evidence that those extra sales are profitable.

Thursday's report will therefore be useful even without a separate robotics profit figure. I want to hear whether unit costs are improving, whether benefits extend beyond the newest facilities and whether efficiency is helping Amazon absorb growth without a matching increase in resources. Several quarters of consistent evidence would be more persuasive than one strong margin number.

Much of the debate over AI's payoff has centered on what companies spend on it. Amazon is a case in point: Its free cash flow swung to an outflow of $7.6 billion over the 12 months through June, driven by a $66.1 billion increase in spending on property and equipment that the company attributed primarily to AI. Its warehouses offer a different test; the customers, orders and delivery network already exist, so any gain has to come from making the trip to the customer's door cheaper.

For shareholders, the decisive measure is how much more cash Amazon can earn from the orders it already knows how to win.

-Jurica Dujmovic

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

10-28-26 0630ET