Amazon’s $200 Billion AI Bet: From Corporate Cuts to Warehouse Robots

Amazon just reported record quarterly revenue of $213.4 billion. Wall Street’s reaction? The stock dropped 5% since the announcement.

The reason isn’t what the company earned today, it’s what it’s spending tomorrow: $200 billion on capital expenditures in 2026, the largest infrastructure investment in Amazon’s history. 

This is Amazon's classic infrastructure playbook: sacrifice short-term margins to build long-term competitive moats that rivals can't replicate. The company has run this strategy before, warehouse networks in the 1990s (years of losses), AWS (launched 2006 and unprofitable until 2015, now generating $45.6 billion in operating income) and Kiva robotics (acquired 2012 for $775 million, now having deployed 1 million units). 

The pattern: internalize infrastructure, scale aggressively, then dominate by keeping it proprietary or monetizing it externally. Amazon could even end up offering “Amazon Robotics as a Service,” similar to how AWS provides cloud infrastructure to competitors, creating an entirely new revenue stream.

The Business Model Today: Three Engines Running at Different Speeds

Amazon operates three distinct businesses, each at a different stage of maturity and automation.

1. AWS: The Profit Engine

AWS is Amazon’s cash machine, generating $35.6 billion in Q4 2025 revenue, up 24% year-over-year. Operating margins hover around 35%, and the segment carries a backlog of $244 billion in contracted revenue, up 40% from last year. That backlog represents future revenue already locked in through multi-year contracts with enterprise customers: $244 billion worth of cloud services that AWS will deliver and recognise as revenue over the coming years.

Highly automated, capital-intensive business, data centres don’t need warehouses full of workers picking items off shelves and only require engineers managing servers and AI systems.

The segment is also powering automation elsewhere. AWS custom chips now generate over $10 billion in annual revenue and are growing at triple-digit percentages. These chips power AI systems that are, ironically, replacing human workers across industries.

AWS is what Amazon’s other businesses aspire to become: high-margin, scalable, and largely automated.

2. Advertising: The Growth Engine

Amazon’s advertising business brought in $21.3 billion in Q4, up 22% year-over-year. This segment operates at high margins with relatively low capital intensity. Automated ad targeting systems, powered by AI, match sellers with customers without requiring much human intervention.

While not as automated as AWS, advertising represents another high-margin business that scales without proportional increases in headcount.

3. E-commerce: The Transformation Target

E-commerce remains Amazon’s largest business by revenue but carries the lowest margins. This is where 46,000 job cuts are concentrated, where $200 billion capital expenditure is planned and where robotic deployment is focused. Amazon is transforming its retail operations from a labour-intensive model, to a capital-intensive model powered by robotics and AI. It’s similar to how Tesla builds highly automated factories: massive capital investment upfront, but each car (or in Amazon’s case, each package) costs less to produce once the robots are operational.

The $200 Billion Equation: AI Everywhere

The $200 billion planned spend is funding a three-front transformation: AI eliminating corporate jobs today, AI powering AWS growth tomorrow, and robotics rebuilding warehouse operations for the next decade.

AI Eliminating Corporate Roles (Happening Now)

Between October 2025 and February 2026, Amazon eliminated 46,000 corporate jobs across HR, operations, devices, and even AWS teams, representing roughly $4 to $6 billion in annual operating expenses.

CEO Andy Jassy stated explicitly in June 2025 that increased use of AI tools would lead to job cuts by automating repetitive corporate tasks including HR workflows and employee enquiries, operations analytics and reporting, code review and testing via AWS agents, and customer service routing.

AI Powering AWS Growth (The Revenue Engine)

While AI eliminates corporate jobs, it's simultaneously driving AWS revenue growth. Amazon’s Trainium chips train and run AI models at lower cost. 1.4 million Trainium2 chips have been deployed and are powering the majority of AI workloads on Amazon's cloud platform.

Then there's Project Rainier, the world's largest AI compute cluster, housing over 500,000 Trainium2 chips in a single facility. Anthropic (the company behind the AI assistant Claude) uses it to train its latest models. Trainium3 and Trainium4 chips promise 6x compute performance improvements.

On the software side, Amazon Bedrocknow offers 20+ fully managed models (from providers including Anthropic, Google, OpenAI, and Amazon's own Nova family) used by over 100,000 companies.The AWS segment generated $35.6 billion in Q4 revenue (up 24% year-over-year) with 35% operating margins and holds a $244 billion backlog (up 40%), representing contracted future revenue already locked in. The AI infrastructure spending is building the compute capacity to power this enterprise adoption wave.

Robotics Transforming Warehouses (The Long Play)

While corporate AI cuts deliver near-term savings, the robotics transformation is the multi-decade bet.

Amazon has deployed over one million robots across 300 fulfilment centres worldwide, including Hercules robots (lifting up to 1,250 pounds of inventory), Pegasus robots (handling individual packages on precision conveyor belts), Proteus (Amazon’s first fully autonomous mobile robot, navigating around human workers), and Sparrow (an AI-powered robotic arm for picking individual items).

According to internal Amazon documents reported by The New York Times, the company plans to replace more than 600,000 jobs by 2033 and automate 75% of its warehouse operations.

Morgan Stanley analysts estimate this automation could save Amazon up to $4 billion annually by 2027 alone.

The breakthrough isn’t just deploying more robots; it’s using AI to coordinate them. Amazon recently launched DeepFleet, a generative AI foundation model that functions as an intelligent traffic management system for the entire robot fleet, coordinating movements to optimise navigation across fulfilment centres.The result: a 10% improvement in robot travel time.

The economics shift from variable to fixed costs. Warehouse workers represent ongoing labour expenses that scale with headcount and wage inflation. Robotic infrastructure requires upfront capital but delivers predictable costs through depreciation over 5 to 10 years.

Mixed Messages

Notably, some leaked documents reveal Amazon’s sensitivity around public perception recommending avoiding terms like “automation” and “AI” in favour of “advanced technology,” and replacing “robot” with “cobot” to emphasise collaboration between humans and machines.

Amazon’s public statements emphasise job creation and upskilling. The company highlights its Career Choice programme, which has trained over 700,000 employees since 2019, and notes that its next-generation fulfilment centres require 30% more employees in reliability, maintenance, and engineering roles.

An Amazon spokesperson responded to the leaked documents: “Our investments will continue to create substantial employment, emphasising higher-paying positions… efficiency gains in one area enable us to invest in other areas that create additional value for customers.”

Yet the internal target of 75% warehouse automation by 2033 suggests a fundamental shift in workforce composition over the next decade.

What the Earnings Actually Reveal

Amazon’s Q4 2025 results show strong fundamentals despite the shock capital expenditure plans

Revenue grew 12% year-over-year to $213.4 billion. Operating income increased to $25 billion, though that includes $2.4 billion in special charges including $730 million in estimated severance costs and $610 million in asset impairments related to physical stores. 

Here’s what the segment performance reveals: