What Management Said
Read the full Q3 2025 transcript ↗Our guidance incorporates the order trends that we've seen to date and what we believe today to be appropriate assumptions. Our guidance assumes, among other things, that we don't conclude any additional business acquisitions, restructurings, or legal settlements. It's not possible to accurately predict demand for our goods and services, and therefore our actual results could differ materially from our guidance. We saw strong growth across our business in Q3, and we're reporting $180.2 billion in revenue, up 12% year-over-year, excluding the impact from foreign exchange rates.
AWS is growing at a pace we haven't seen since 2022, re-accelerating to 20.2% year-over-year, our largest growth rate in 11 quarters. Backlog grew to $200 billion by Q3 quarter end and doesn't include several unannounced new deals in October, which together are more than our total deal volume for all of Q3. SageMaker makes it much simpler for companies to build and deploy their own foundation models. Bedrock gives customers leading selection of foundation models and superior price performance to deploy inference into their next-generation applications.
It's why we launched AgentCore, a set of infrastructure building blocks that allow builders to deploy secure, scalable agents. Cohere Health is using AgentCore to deploy agents that will reduce medical review times by up to 30%-40%. Anthropic is using it now to build and deploy its industry-leading AI model, Claude, which we expect to be on more than 1 million Trainium2 chips by year-end. Trainium2 continues to see strong adoption, is fully subscribed, and is now a multi-billion-dollar business that grew 150% quarter-over-quarter.
What went well
- AWS re-accelerated to 20.2% YoY growth (revenue $33B), its fastest in 11 quarters and up 270bps QoQ, reaching a $132B annualized run rate.
- AWS operating income was $11.4B, up $2.1B QoQ, driven by growth plus efficiency gains.
- Excluding two special charges, worldwide operating income would have been $21.7B, $1.2B above the high end of guidance.
- Advertising revenue accelerated for a third consecutive quarter to $17.7B.
- Backlog grew to $200B and did not yet include several unannounced October deals that together exceeded all of Q3's deal volume.
- Trainium2 became a fully subscribed multi-billion-dollar business, up 150% QoQ; Project Rainier came online with ~500,000 Trainium2 chips (heading to 1M by year-end).
- Added more than 3.8 GW of power over the trailing 12 months (more than any other cloud provider); ~1 GW+ more expected in Q4.
- Rufus reached 250M active customers with monthly users +140% YoY and is on track for $10B+ in incremental annualized sales; committed $4B+ to expand rural delivery.
What went wrong
- Two special charges cut operating income by $4.3B: a $2.5B FTC legal settlement (North America) and $1.8B in severance for role eliminations (all three segments), pulling reported worldwide operating income down to $17.4B.
- North America operating margin was just 4.5% (op income $4.8B); excluding the FTC charge it would have been 6.9% ($7.3B).
- International operating margin was 2.9% (op income $1.2B), pressured by the severance charge (margins expanded YoY excluding it).
- Trailing 12-month free cash flow fell to $14.8B; management flagged AWS margins will keep fluctuating given the depreciation from heavy AI CapEx.
- Trainium at scale was still limited to a small number of very large customers, with broader availability deferred to Trainium3.
Guidance Changes
| Metric | Period | Current guidance |
|---|---|---|
| Full-year cash CapEx | FY 2025 | ~$125B (expected to increase in 2026) |
| AWS operating margins | Ongoing | Expected to keep fluctuating with the level of investment / depreciation |
| Delivery speed | FY 2025 | On track for fastest-ever Prime speeds for a third consecutive year |
Performance Breakdown
| Metric | YoY | Note |
|---|---|---|
| Worldwide revenue | +12% ex-FX ($180.2B) | Strong growth across all segments; 90bps favorable FX; worldwide paid units +11%. |
| Worldwide operating income | n/a ($17.4B reported; $21.7B ex-charges) | $4.3B of special charges ($2.5B FTC settlement + $1.8B severance); ex-charges $1.2B above guidance. |
| AWS revenue | +20.2% ($33B) | Re-acceleration of 270bps QoQ on strong AI and core-services growth plus new capacity coming online; $132B run rate, largest growth in 11 quarters. |
| AWS operating income | n/a ($11.4B) | Continued growth plus efficiency focus, partly offset by higher depreciation from Gen-AI data-center buildout. |
| North America operating income | n/a ($4.8B reported; $7.3B ex-FTC) | Reported margin 4.5% (6.9% ex the $2.5B FTC charge); also carried part of the severance charge. |
| International operating income | n/a ($1.2B) | Margin 2.9%; excluding severance, margins expanded YoY on transportation productivity. |
| Advertising revenue | growth accelerated 3rd consecutive quarter ($17.7B) | Full-funnel approach resonating on an increasingly large base. |
| Cash CapEx | n/a ($34.2B; $89.9B YTD) | Primarily AWS AI and core services plus custom silicon (Trainium) and fulfillment/transportation. |
| Net income | n/a ($21.2B) | Includes a $9.5B pre-tax non-operating gain on the Anthropic investment. |
Earnings Call Themes & Trends
| Topic | Previous mention | Current period | Trend |
|---|---|---|---|
| AWS growth trajectory | 17.5% YoY in Q2 2025 | 20.2% YoY in Q3 2025 (fastest in 11 quarters), $132B run rate | — |
| AWS capacity / power | Power the biggest constraint (Q2) | Added 3.8 GW over trailing 12 months, on track to double capacity by 2027; 1 GW+ more in Q4 | — |
| Trainium2 | Backbone of Anthropic/Bedrock, ramping (Q2) | Fully subscribed, +150% QoQ; Project Rainier online with ~500K chips heading to 1M | — |
| Special charges | None flagged in Q2 | $4.3B of charges ($2.5B FTC settlement + $1.8B severance) hit reported operating income | — |
| Grocery / perishables | Perishables pilot expanding (Q2) | Same-day perishables in 1,000+ cities, en route to 2,300 by year-end; grocery GMS $100B+ TTM | — |
| Headcount / culture | Not a focus in Q2 | Role eliminations described as culture-driven (removing layers), not AI- or cost-driven | — |
Q&A Summary
How capacity-constrained are you, and what Trainium demand are you seeing beyond your major customers? (Justin Post, Bank of America)
Jassy: added 3.8 GW in the last year with 1 GW+ more in Q4 and plans to double capacity by end of 2027; monetizing capacity as fast as it comes in; power is today's bottleneck, may shift to chips. Trainium2 is fully subscribed, a multi-billion-dollar business up 150% QoQ; Trainium3 previews at year-end with fuller volumes in early 2026 and strong large- and mid-size customer interest.
What hurdles must Trainium3 clear to broaden adoption versus third-party chips? (Brian Nowak, Morgan Stanley)
Jassy: AWS will always offer multiple chip options and buys a lot of NVIDIA (unconstrained). Trainium3 is expected ~40% better than Trainium2 (already ~30-40% better price/performance than GPUs); AWS must deliver volume quickly, keep improving the software ecosystem, and build credibility via proof points like Project Rainier.
How is Project Rainier architected/differentiated and will it expand beyond Anthropic? (Doug Anmuth, JPMorgan)
Jassy: Rainier is a very large 500,000+ (heading to 1M) Trainium2 cluster, hard to build at scale, combining AWS infrastructure heritage with chip price/performance; it is specific to Anthropic, but other customers want large Trainium clusters and will get the chance with Trainium3.
Does same-day perishables scale mean you no longer need Amazon Fresh stores, and how will headcount trend with AI efficiencies? (Mark Mahaney, Evercore)
Jassy: grocery GMS is $100B+ TTM (top-three U.S. grocer) excluding Whole Foods/Fresh; perishables same-day (1,000 cities, 2,300 by year-end) is changing the weekly stock-up while still experimenting with formats like Daily Shop. The recent role eliminations were culture-driven (removing layers, more ownership), not financially or AI-driven right now.
How big is the robotics/physical-AI opportunity as an efficiency and reinvestment driver? (Eric Sheridan, Goldman Sachs)
Jassy: over 1 million robots deployed with much more invention in flight; robots improve safety, productivity and speed while humans remain central, enabling continued customer-experience and cost improvements.
How does Amazon think about agentic commerce going forward? (John Blackledge, TD Cowen)
Jassy: bullish long term; AI will make online discovery better than physical stores. Amazon has Rufus and Buy for Me and is talking with third-party agents, but today their experience is poor (no personalization/history, wrong prices and delivery estimates); agentic commerce should expand online shopping and benefit Amazon given selection, value and delivery.
How much of AWS acceleration is core vs AI, and can you disaggregate advertising (core vs DSP vs Prime Video)? (Colin Sebastian, Baird)
Jassy: growth is broad across AI (inference, training, Trainium, Bedrock, SageMaker) and core migrations, with AgentCore/Strands unlocking agent builds; believes AWS can sustain this clip for a while. Every advertising offering grew meaningfully thanks to a full-funnel offering; stores, video and DSP (boosted by Roku, Netflix, Spotify, SiriusXM integrations) are all growth vectors.
More on Amazon Com Inc
Reported 2025-10-30 · figures from the Amazon Com Inc Q3 2025 earnings call.
See how Top Bucket AI works for your firm
Request DemoTop Bucket AI Research
Stay ahead of private markets
Research and market intelligence for private-markets professionals.
✓
You're subscribed.
Thanks for signing up.