Amazon.com, Inc. earnings call
Capex plan jumps to about $200B, mostly AWS, monetized immediately
Amazon's Q4 2025 beat expectations with 24% AWS growth and announced record-breaking $200B capex plan, highlighting aggressive investment in AI infrastructure to meet surging demand, while also expanding LEO satellite and quick-commerce initiatives. Management exuded high confidence in ROI on AI spending despite market uncertainty on the 3-year payback horizon. Reported Q4 revenue of $213.4B (+12% YoY), operating income of $25B, and free cash flow of $11.2B.
Buzzberg read Capex plan jumps to about $200B, mostly AWS, monetized immediately Amazon's Q4 2025 beat expectations with 24% AWS growth and announced record-breaking $200B capex plan, highlighting aggressive investment in AI infrastructure to meet surging demand, while also expanding LEO satellite and quick-commerce initiatives. Management exuded high confidence in ROI on AI spending despite market uncertainty on the 3-year payback horizon. Reported Q4 revenue of $213.4B (+12% YoY), operating income of $25B, and free cash flow of $11.2B. Read full analysisCollapse analysis
Amazon's Q4 2025 beat expectations with 24% AWS growth and announced record-breaking $200B capex plan, highlighting aggressive investment in AI infrastructure to meet surging demand, while also expanding LEO satellite and quick-commerce initiatives. Management exuded high confidence in ROI on AI spending despite market uncertainty on the 3-year payback horizon. Reported Q4 revenue of $213.4B (+12% YoY), operating income of $25B, and free cash flow of $11.2B.
- AWS revenue accelerated to $35.6B, growing 24% YoY, aided by both core cloud and AI services.
- Announced 2026 capex plan of $200B, focused on AWS and AI infrastructure, signaling massive growth investment.
- Revealed $244B AWS backlog, providing strong multi-year demand visibility.
What matters now
The highest-signal changes from the call.
AWS backlog hits $244B, up 40% year-over-year
Actuals
| Metric | Reported | Change |
|---|---|---|
| AWS Revenue | $35.6B | Reported |
| Revenue | $213.386B | Reported |
| EPS | $1.95 | Reported |
| Gross margin | 48.47% | Reported |
| Operating margin | 11.71% | Reported |
| Free cash flow | $14.937B | Reported |
Forward guidance
| Metric | Period | Range | Midpoint | Status |
|---|---|---|---|---|
| Capex | FY2026 | $200B | $200B | Initiated |
| Operating margin | FY2026 Q1 | $16.5B–$21.5B | $19B | Maintained |
| Revenue | FY2026 Q1 | $173.5B–$178.5B | $176B | Maintained |
Management read
Bullish
Andy repeatedly called AI an 'unusual opportunity,' cited accelerating AWS momentum and strong demand signals, and expressed high confidence in strong returns on invested capital.
Management AI read
Management framed AI as an unusual, broad opportunity: AWS is seeing demand across AI labs, productivity workloads, and upcoming enterprise production applications, which they expect to be the largest and most durable pool. They highlighted agentic AI as the main way companies will capture value, and said custom Trainium silicon plus services like Bedrock and AgentCorp give AWS a price-performance
Investment and capacity
Capex is being raised sharply to roughly $200 billion, predominantly for AWS, because demand is high and capacity is monetized immediately. Management highlighted record capacity additions, said they plan to double power again by end-2027, and noted LEO satellite costs will pressure near-term results before later capitalization.
Companiesreturns since call
Customers
HSBC is a new AWS customer for cloud services.
Evidence
“Since our last call, we announced new agreements with OpenAI, Visa, MBA, BlackRock, Perplexity, Lyft, United Airlines, DoorDash, Salesforce, U.S. Air Force, Adobe...”
Anthropic is a major AI customer using Amazon's custom Tranium chips for training its next model, indicating a deepening strategic partnership and substantial capacity commitment.
Evidence
“Anthropic is building their next, they're training the next cloud model on top of Tranium 2.0. And that's what Project Rainier is. We talked about 500,000 chips there.”
OpenAI is a significant customer and partner for AWS, with a large agreement announced in November 2025 that is expected to deepen over time.
Evidence
“We announced an agreement with OpenAI in November. We're excited about that agreement. It's a big one. We have a lot of respect for the company, and we hope to continue to extend our partnership over time.”
Suno is a notable AWS customer using AI for music generation, mentioned as a example of enterprise and AI-native workload on AWS.
Evidence
“Suno is a really cool example of an AWS customer that's kind of reinvented how you can write music and build music.”
Partners
Competitors
Amazon's custom Graviton and Tranium chips combined are now over a $10 billion annualized run-rate business, highlighting the scale of its silicon efforts. — Amazon's move of critical workloads to its own silicon is reducing its dependency on Nvidia, potentially impacting Nvidia's pricing power and market share in cloud data centers.
Evidence
“Tranium-2 is 30% to 40% more price performance than comparable GPUs.”
Supply-chain alpha · 4returns since call
Amazon's AWS backlog is $244 billion, up 40% year-over-year, providing significant revenue visibility beyond the annualized run rate.
Evidence
“Our backlog is $244 billion. That's up 40% year over year. I think it's up 22% quarter over quarter.”
All of Amazon's Tranium 3 supply is expected to be committed by mid-2026, indicating demand for custom silicon is outstripping supply.
Evidence
“We're seeing very strong demand for Tranium 3 and expect nearly all of our Tranium 3 supply of chips to be committed by mid-2026.”
AWS added 1.2 gigawatts of data center power in Q4 alone, and is doubling total power capacity by the end of 2027.
Amazon's custom Graviton and Tranium chips combined are now over a $10 billion annualized run-rate business, highlighting the scale of its silicon efforts.
Methodology & coverage
Management-only analysis. All 19 validated company mentions are shown. Reported actuals and forward guidance are kept separate. Public evidence is limited to eight short attributed quotes. AI-generated analysis can be incomplete or wrong; verify important claims against the original source.