tapebrief
AMZN · Q3 2025
AMZN · Q1 2026

AMZN · Q4 2025 Earnings

Bullish

Amazon

Reported February 5, 2026

30-second summary

AWS grew 24% YoY to $35.6B — a further 380bps reacceleration from Q3's 20.2% — while Q4 revenue of $213.4B (+14% YoY) printed at the absolute top of guidance. The signal that matters most: management put a $200B FY2026 capex number on the page, roughly 5–6x historical levels, and paired it with a Q1 operating income guide ($16.5–21.5B) that runs ~$6.0B below Q4 actual at midpoint. AWS reaccelerating two quarters in a row while AWS operating margin holds sequentially at 35.0% (despite compressing 190bps YoY from 36.9%) is the bull case; the $200B capex commitment with no disclosed financial guardrails is what the bear case will anchor on.

Headline numbers

EPS

Q4 FY2025

$1.95

Revenue

Q4 FY2025

$213.40B

+14.0% YoY

Operating margin

Q4 FY2025

11.7%

Key financials

Q4 FY2025
MetricQ4 FY2025YoYQ3 FY2025QoQ
Revenue$213.40B+14.0%$180.20B+18.4%
EPS$1.95—$1.95+0.0%
Operating margin11.7%—9.7%+200bps

Guidance

Q4 FY2025 revenue beat guidance with 14% YoY growth; Q1 FY2026 guided with 11-15% growth and $200B annual CapEx commitment reflecting accelerated AI and infrastructure investments.

Guidance is issued one quarter forward. The Prior-guide column references the guide issued last quarter for the period just reported; the New-guide column is for next quarter.

Actuals vs prior guidance

MetricPeriodPrior guideActualΔResult
RevenueQ4 FY2025$206.0 billion to $213.0 billion$213.4 billion+$0.4 billion above high endBeat
Operating IncomeQ4 FY2025$21.0 billion to $26.0 billion$24.9 billion-$1.1 billion below midpoint ($25B midpoint)Missed

New guidance

MetricPeriodGuideYoY
RevenueQ1 FY2026$173.5 billion to $178.5 billion11% to 15%
Operating IncomeQ1 FY2026$16.5 billion to $21.5 billion—
Capital ExpendituresFY 2026~$200 billion—

Segment performance

Q4 FY2025
SegmentQ4 FY2025YoY
North America$127.1B+10.0%
International$50.7B+17.0%
AWS$35.6B+24.0%
Online stores$83B+10.0%
Third-party seller services$52.8B+11.0%
Advertising services$21.3B+23.0%
Subscription services$13.1B+14.0%
AWS Revenue Growth (YoY)24%—
Advertising Services Growth (YoY)23%—

Platform metrics

Q4 FY2025
SegmentQ4 FY2025
WW Paid Units Growth (YoY)12%
AWS Custom Chips Annual Revenue Run Rate$10B+

Profitability

Q4 FY2025
SegmentQ4 FY2025
AWS Operating Margin35.0%
North America Operating Margin9.0%
International Operating Margin2.1%
Operating Cash Flow (TTM)$139.5B

Management tone

Narrative arc: Q2 "demand exceeds supply" → Q3 "AWS reaccelerated, AI is revenue" → Q4 "we are spending $200B to own the AI infrastructure stack."

The capex framing pivoted from "investing to meet demand" to "owning the stack." Three quarters ago, management's response to AWS share-loss concerns was the migration tail (85–90% IT spend on-prem). Two quarters ago it was the $195B backlog. This quarter it is the $200B capex commitment and the Trainium 2/3/4/5 roadmap, with Jassy describing AI as "an extraordinarily unusual opportunity" and stating Amazon will "invest aggressively to be a leader." When Mahaney pressed for minimum FCF guardrails during the capex cycle, no specific number was offered. The posture shift is from "trust the backlog conversion" to "trust the multi-year investment thesis" — and from disclosed guardrails to disclosed conviction.

AI demand structure was articulated for the first time as a durable three-segment market. Asked by Sandler about lab concentration, Jassy described demand as "barbelled" — labs (training/inference), enterprise production workloads (which he expects to be "largest and most durable"), and new AI-native applications. The framing is notable because it specifically de-emphasizes lab concentration as a risk, positioning enterprise inference as the long-tail revenue base. Combined with the OpenAI November agreement being described as "big" but explicitly not exclusive ("thousands of companies over time, not a couple"), management is signaling AWS does not need to win the lab war to win the AI revenue war.

Quick commerce and Amazon Leo (Kuiper) moved from optionality to disclosed investment line items. A quarter ago these were peripheral. This quarter management called out "$1 billion of higher year-over-year Amazon Leo costs" baked into the Q1 OI guide, and Jassy walked Nowak through quick-commerce data (Amazon Now in India/UAE/Mexico driving 3x shopping frequency, perishables driving 2x). Both moved from "watch this" to "this is in the operating income guide." The investment narrative is broader than AWS alone.

Headcount direction stayed undisclosed; productivity narrative shifted to robotics. Q3's ~18,000 role reduction was framed as culture/structure. Q4 made no further headcount disclosure but Jassy used the Nowak retail question to emphasize 1M+ robots deployed with continued expansion — the productivity story management wants investors to anchor on for retail is robotics, not headcount.

Q&A highlights

Mark Mahaney · Evercore ISI

Request for insight into strong long-term return on invested capital claims, specifically asking about CapEx cycle duration, profitability expectations, and minimum free cash flow levels management won't go below during the CapEx cycle.

Brian Osowski and Andy Jassy provided detailed confidence in returns, citing AWS 35% operating margin (up 40 bps YoY), immediate monetization of capacity as installed, strong demand signals, and historical AWS track record. Andy emphasized capacity is being monetized immediately, that AI represents unprecedented opportunity, and confidence stems from AWS's proven ability to forecast demand and build efficient data centers. Noted inference will be majority of long-term AI workloads with optimization and price normalization expected over time.

AWS operating margin: 35% in Q4, up 40 basis points YoYCapacity monetized as fast as installedAnnualized revenue run rate: $142 billion at 24% YoY growthTranium providing better economics for Amazon alongside customer price benefits

Doug Anmuth · JPMorgan

Asked about Project Rainier with Anthropic after first full quarter, clarification on chip numbers (500k vs 1M mentioned earlier), and whether any financial guardrails or governors exist around spending.

Andy Jassy clarified Tranium 2 is fully subscribed with 500k chips for Project Rainier, with numbers continuing to increase as Anthropic uses more for other workloads. Emphasized Tranium as multi-billion dollar business, Tranium 3 commitment expected by mid-2026 with substantial interest, and Tranium 4 already generating interest for 2027. On guardrails question, Andy reiterated this is 'extraordinarily unusual opportunity' and Amazon will 'invest aggressively' to be leader without disclosing specific financial guardrails.

Tranium 2: fully subscribed, 500k chips deployedTranium multi-billion dollar annualized run rate businessTranium 3: 40% better price performance than Tranium 2; nearly all supply expected committed by mid-2026Tranium 4: coming in 2027 with substantial interest already; Tranium 5 conversations underway

Ross Sandler · Barclays

Asked how AI market concentration among few labs is changing in 2026, and how Amazon might extend OpenAI relationship to support Amazon's own retail and AWS AI efforts.

Andy Jassy described AI demand as 'barbelled' with labs spending heavily on one end and enterprises on the other, with enterprise production workloads in the middle likely becoming largest and most durable segment. Emphasized demand is still in early innings and will grow as AI talent spreads, inference costs drop, and companies achieve success. On OpenAI, stated the November agreement is big and respected but noted this will be thousands of companies over time, not a couple; AWS has significant relationships across many companies and almost every conversation starts with AI.

AI demand structure: labs (training/inference), enterprise production workloads (middle, largest/most durable), and new AI-native applicationsOpenAI agreement announced November; described as 'big'AWS relationships span many AI companies, not concentratedAI movement will involve thousands of companies, not limited to few labs

Michael Morton · Moffitt Nathanson

Asked about how agentic shopping will play out for retail business and potential compression in onsite ads funnel as consumers get better answers from agents.

Andy Jassy expressed optimism about agentic shopping improving customer experience via Rufus, noting 300M customers used Rufus in 2025 and users are 60% more likely to complete purchase. Acknowledged need to work with third-party horizontal agents but highlighted their current limitations (wrong product details, pricing errors, no shopping history). Argued retailers' agents have inherent advantages (selection, pricing, speed, trust) and consumers will likely prefer retailer-owned agents. Noted horizontal agents represent small traffic fraction historically (like early search engines).

Rufus usage: 300 million customers in 2025Rufus users: 60% more likely to complete purchaseThird-party horizontal agents limitations: missing shopping history, inaccurate product details and pricingRetail agent advantages: broad selection, low prices, fast delivery, trust and service

Brian Nowak · Morgan Stanley

Asked to break down both investment areas and efficiency sources expected in global retail business for 2026, including where cost-to-serve savings may come from and where investments (robotics, etc.) will drive durable growth.

Andy Jassy outlined investment areas: selection expansion (luxury and everyday essentials), speed of delivery (key customer differentiator), quick commerce (Amazon Now in India, UAE, Mexico showing strong adoption with 3x frequency increase), and perishables (available same-day in thousands of cities, drives 2x shopping frequency). On efficiencies: regionalization (expanded from 8 to 10 regions, extended to inbound), increased units per box, and robotics (1M+ robots today, more capability planned with safety and productivity benefits). Emphasized both people and robotics will continue in fulfillment network.

Everyday essentials: 1 of 3 units sold; customers tripled shopping frequency in India QuickCommerceAmazon Now (QuickCommerce): 30-minute delivery in India, UAE, Mexico; customers shop 3x more frequentlyPerishables: same-day delivery in thousands of cities; 9 of top 10 items in perishable geographies are perishables; drives 2x shopping frequencyRegionalization: expanded from 8 to 10 regions; extended to inbound

Answers to last quarter's watch list

AWS growth above 20% — AWS printed 24%, a further 380bps reacceleration from Q3's 20.2%. The capacity build is clearly converting to revenue faster than the bear case allowed, and Trainium2 fully subscribed with 1.4M chips landed (500K+ deployed for Project Rainier specifically) corroborates that backlog conversion is structural, not a one-quarter release.
Resolved positively
Q4 operating income relative to $23.5B midpoint — Operating income landed at $25.0B, in the upper half of the $21.0–26.0B guidance band and roughly $1.5B above midpoint. Total margin held at 11.7%; AWS margin held sequentially at 35.0% though it compressed 190bps YoY from 36.9%. AI capex is beginning to weigh on segment margin YoY but has not eaten the sequential line.
Resolved positively
Trainium 3 customer disclosures — Jassy disclosed that Trainium3 will deliver 40% better price/performance than Trainium2 and that "nearly all supply" is expected committed by mid-2026, with Trainium4 generating interest for 2027 and Trainium5 conversations underway. No specific named customers beyond Anthropic were disclosed, but the supply-commitment framing is stronger evidence of broad demand than named logos would have been.
Resolved positively
Connect ARR trajectory beyond $1B — Management did not refresh the Connect agentic AI ARR disclosure this quarter. The lack of a cross-quarter data point is itself notable — when a number was useful as a one-quarter proof point it was disclosed; when sustained growth is harder to assert it is not.
Continue monitoring
Headcount direction post-restructuring — No Q4 headcount disclosure or update on whether further reductions are planned. Management redirected the retail productivity narrative toward robotics (1M+ deployed, continued expansion) rather than headcount as the operating leverage source.
Continue monitoring

What to watch into next quarter

AWS growth sustaining 22%+ — does the 17.5% → 20.2% → 24% trajectory continue, or does the easy-compare period from the 2024 capacity throttle end? A Q1 print of 22%+ would extend the reacceleration; a deceleration back toward 20% would suggest Q4's 24% included a capacity catch-up bolus.

AWS operating margin trajectory — Q4 held at 35.0% sequentially but compressed 190bps YoY from 36.9%. Watch whether the $200B FY2026 spend compresses segment margin further in Q1, or whether the chip-led cost advantage (Trainium economics for Amazon and customers) stabilizes the line.

Q1 operating income inside the $16.5–21.5B band — the upper half ($19–21.5B) would confirm Leo and quick-commerce investment is paced; the lower half would signal investment is front-loaded and 2026 will be a margin-pressure year.

Trainium 3 customer logos — Jassy said "nearly all supply" of Trainium3 will be committed by mid-2026. Watch whether any named non-Anthropic customer is disclosed on the Q1 print as evidence the chip stack is scaling beyond a single anchor tenant.

FY2026 capex variance from $200B — first quarter where management could raise or trim the FY2026 capex number. A raise would intensify the guardrails debate; a trim would signal supply constraints are pacing the build.

International operating margin against Q4's 2.1% — management explicitly flagged "sharper prices" as a 2026 investment. Watch whether International margin compresses below 2% in Q1 as a marker of how aggressive the international reinvestment actually is.

Sources

  1. Amazon Q4 FY2025 earnings press release (SEC filing): https://www.sec.gov/Archives/edgar/data/1018724/000101872426000002/amzn-20251231xex991.htm
  2. Q4 FY2025 earnings call Q&A.
  3. Tapebrief Q3 FY2025 and Q2 FY2025 AMZN briefs (for cross-quarter context).

Get the next brief, free.

We publish analyst-grade earnings briefs the same day or morning after every call — headline numbers, segment KPIs, Q&A highlights, and tone analysis. Free during beta.

This is not investment advice.