Date: 2026-04-01 Quarter: Q4 FY25 (Dec-2025) | Reported: Feb 5, 2026 Market cap: ~2, 250B|Shareprice: 208 | EV: ~$2,193B FY25 Revenue: $716.9B | FY25 EPS: $7.17 | TTM FCF: $11.2B
I haven't covered Amazon in the newsletter as a primary recommendation — at $2.25 trillion in market capitalization, it sits outside the mid-cap growth universe where I spend most of my analytical effort. But applying my framework, Amazon just delivered what I consider to be one of the more impressive quarters from any company in the past year. AWS re-accelerated to 24% year-over-year growth on a $142 billion annualized revenue run rate — the fastest growth in 13 quarters — and the company posted $213.4 billion in total revenue, beating the top end of its guidance range. Adjusted for $2.4 billion in special charges, operating income reached $27.4 billion. The complication, and the reason the stock trades at what I consider a meaningful discount to intrinsic value, is the $200 billion FY26 capex guidance with no FCF floor — a point two analysts pressed on the call and management politely declined to address. At a GAAP P/E of approximately 29X against 30% earnings growth, the PEG ratio is essentially 1.0X — fair value territory on a consolidated basis, but the sum-of-parts math suggests the market is not appropriately valuing the high-growth, high-margin segments within the conglomerate. For holders, this is a strong hold. For those building positions in AI infrastructure, Amazon deserves serious consideration alongside the pure-plays.
Headline KPIs:
The Q4 revenue print of $213.4 billion deserves disaggregation, because the consolidated 13.6% YoY growth rate understates what is actually happening inside this business.
AWS grew 24% YoY to $35.6 billion, adding $7 billion year-over-year and $2.6 billion sequentially. The sequential acceleration pattern is striking: $1.5B -> $1.6B -> $2.1B -> $2.6B over the past four quarters. This is not a one-quarter blip — it is a sustained inflection driven by AI workload conversion from the 244billionbacklogthatgrew402.6B) than most of the companies I cover generate in total revenue.
Advertising grew 23% to $21.3 billion, making it a $69 billion full-year business. This is Amazon's highest-margin revenue line (margins are undisclosed but likely 50%+ operating margin given the near-zero marginal cost structure) and it has grown at 20%+ for eight consecutive quarters. Prime Video's ad-supported audience reached 315 million viewers globally, up from 200 million in early 2024 — a 58% expansion of the advertising supply surface.
North America retail grew 10% YoY with a record 9.0% operating margin. Paid units grew 12% — the highest quarterly rate in 2025. Third-party seller services grew 11% to $52.8 billion. The retail business is not exciting from a growth perspective, but it is generating significant operating leverage that subsidizes the AI infrastructure buildout.
International grew 17% YoY (11% ex-FX) to $50.7 billion. Operating margin compressed to 2.1% due to $1.1 billion in Italy tax settlement and lawsuit charges, plus deliberate investment in quick commerce and sharper pricing. Excluding charges, margin expanded YoY — but this segment remains a drag on consolidated profitability.
The revenue beat of 1.85% versus guidance midpoint extends the consecutive beat streak to seven quarters. Amazon consistently beats the high end of its own range. I would characterize Amazon's guidance practice as "deliberately undemanding estimates" — the same pattern I observe at my highest-conviction holdings.
The margin story at Amazon requires careful parsing to separate the underlying business trajectory from the capex investment cycle.
Gross margin: 48.5% GAAP, up 120bps YoY. The trend from 42.6% (Q4 FY22) to 48.5% (Q4 FY25) represents a fundamental improvement in Amazon's revenue mix — higher-margin AWS, advertising, and subscription revenue are growing faster than lower-margin retail, structurally lifting the blended margin.
Operating margin: 11.7% GAAP (12.8% adjusted for $2.4B charges). This compares to 1.8% in Q4 FY22 — a transformation from near-zero profitability to a genuinely profitable business. By segment:
| Segment | Q4 FY25 Op Margin | Q4 FY24 | YoY Change |
|---|---|---|---|
| North America | 9.0% | 8.0% | +100bps (record) |
| International | 2.1%* | 3.0% | -90bps (charges) |
| AWS | 35.0% | 36.9% | -190bps |
*International includes $1.1B charges; ex-charges margin expanded YoY.
The AWS margin decline from the 39.5% Q1 FY25 peak to 35.0% is worth monitoring. This is 450bps of compression in three quarters, and the CFO's language — "will fluctuate" — is deliberately vague. I suspect the new AI infrastructure is carrying higher depreciation loads before it is fully utilized. The critical question is whether AWS margins stabilize at 33-35% or continue compressing toward the low 30s as $200 billion in capex flows through.
FCF margin: TTM FCF margin of 1.6% ($11.2B on $716.9B revenue) is optically atrocious. But this metric is entirely a function of the capex investment. Operating cash flow of $139.5 billion (19.5% margin) grew 20% YoY and demonstrates that the underlying business generates enormous cash. The delta between operating CF margin (19.5%) and FCF margin (1.6%) is 17.9 percentage points — all of it capex.
SBC leverage: Stock-based compensation declined to 2.1% of revenue from 2.7% a year ago. In absolute terms, SBC fell 12% YoY to $4.4 billion even as revenue grew 13.6%. This is meaningful operating leverage.
AWS backlog: $244 billion (+40% YoY, +22% QoQ). This is the single most important data point in the release. The backlog is growing at nearly 2x the rate of AWS revenue (40% vs. 24%), which creates a widening bullish divergence — a textbook leading indicator that AWS growth will sustain in the mid-20s or accelerate further. At current run rate, the backlog provides 1.7 years of revenue coverage, up from approximately 1.4 years a year ago.
Custom chips: >$10B combined ARR, triple-digit YoY growth. Trainium and Graviton together have reached the scale of a significant standalone company. Trainium is "fully subscribed" and offers 30-40% better price-performance than comparable GPUs — creating both a revenue accelerant and a structural margin advantage. Trainium3 is in production workloads with "nearly all supply committed by mid-2026." Trainium4 expected in 2027 with 6x the FP4 compute.
Bedrock: multibillion-dollar ARR, +60% QoQ customer spend growth. This is explosive adoption of Amazon's AI model platform. At 60% quarterly growth, this business is compounding faster than any other segment I track.
Amazon Connect: $1B ARR, 30%+ growth. An enterprise software business within AWS that confirms platform extensibility beyond core compute.
Q1 FY26 guidance: 173.5B−178.5B (+13% YoY at midpoint). Given the seven-quarter beat cadence, I expect actual revenue closer to 179−181 billion. The operating income range of 16.5B−21.5B ($19B midpoint) implies modest growth over Q1 FY25's 18.4B—widerrangeaccountsfor 1B in Amazon Leo (Kuiper satellite) incremental costs and international quick commerce investment.
Cloud positioning: AWS maintains approximately 32% global cloud infrastructure market share per Synergy Research — well ahead of Azure at ~23% and Google Cloud at ~11%. Azure is growing faster in percentage terms (31% reported vs. AWS 24%), but on a smaller revenue base. The absolute dollar gap between AWS and Azure continues to widen — AWS added approximately $21 billion in incremental revenue in FY25, versus Azure's estimated $17-18 billion. More critically for the investment thesis, AWS's custom silicon strategy (Trainium, Graviton) creates cost structure advantages that neither Azure nor GCP can easily replicate.
Jassy's framing on the call was instructive: "Very different having 24% on $142 billion versus a higher percentage on a smaller base." He is correct. Growing $142 billion in annualized revenue at 24% means adding $34 billion in new run-rate AWS revenue this year — larger than Snowflake, Datadog, and CrowdStrike combined.
AI stack breadth: Amazon's AI competitive advantage is not just compute — it is the full stack. Bedrock (models), SageMaker (training/fine-tuning), Nova Forge (custom enterprise models), Strands (agents), AgentCore (agent infrastructure), and custom chips. This is the most complete AI stack of any cloud provider. The customer list announced since last earnings — OpenAI, Visa, NBA, BlackRock, Salesforce, CrowdStrike, Adobe — reads like a directory of the Fortune 100.
Retail moat: Fastest delivery speeds for the third consecutive year while simultaneously reducing cost to serve. Same-day delivery reached ~100 million US customers. The "Add to Delivery" feature (10% of Prime volumes in 6 months) is a logistics innovation that further embeds Amazon in daily shopping habits. Rufus AI assistant has been used by 300M+ customers with ~$12B in incremental annualized sales attributed to it.
New vectors: Kuiper (satellite broadband) is a genuine optionality asset — 180 satellites launched, commercial launch in 2026, enterprise terminal with 1Gbps speeds, contracts signed with AT&T, DIRECTV, JetBlue. Alexa+ is available to all US customers. These may not move the needle in FY26 but represent free optionality on a $2.25 trillion market cap.
This is where I apply my framework most rigorously. Amazon's valuation requires segment-level thinking because the consolidated multiples are misleading.
Consolidated metrics:
| Metric | Value | Assessment |
|---|---|---|
| P/E (GAAP TTM) | 29x | Fair — on 30% EPS growth = PEG 0.97 |
| EV/Revenue (TTM) | 3.1x | Meaningless at consolidated level (retail drag) |
| P/Operating CF | 16.1x | Attractive — $139.5B OCF growing 20% |
| EV/FCF | 196x | Distorted by capex — not investable metric currently |
| PEG (P/E / EPS growth) | 0.97x | Fair value threshold; attractive vs. MSFT at ~1.5x PEG |
The consolidated P/E of 29x on 30% EPS growth produces a PEG just below 1.0x. By my valuation benchmarks for profitable companies growing 15-20% at the consolidated level (which is where Amazon falls given 12-14% revenue growth translating to 25-30% earnings growth via margin expansion), a PEG of 1.0x is fair to slightly attractive. Not a screaming buy, but not a premium either.
The sum-of-parts tells a more compelling story:
| Segment | Revenue Base | Growth | Comp Multiple | Implied Value |
|---|---|---|---|---|
| AWS | $142B ARR | 24% | 12x EV/Rev | $1,704B |
| Advertising | $69B FY25 | 23% | 7x EV/Rev | $483B |
| N. America Retail | $426B FY25 | 10% | 0.5x EV/Rev | $213B |
| International | $162B FY25 | 13% | 0.3x EV/Rev | $49B |
| Subscriptions | $50B FY25 | 14% | 2.5x EV/Rev | $125B |
| Kuiper / Other | -- | -- | -- | $30B |
| Total enterprise value | ~$2,604B | |||
| Plus: net cash | +$57B | |||
| Implied equity | ~$2,661B |
I am deliberately using conservative multiples here — AWS at 12x is below the cloud cohort average for 24% growers, which typically commands 8-14x EV/S (I use the midpoint). For reference, a 20-30% CAGR cloud company in the current market trades at 4-8x EV/S for the typical name — AWS deserves a premium for scale, profitability (35% op margin), and the $244B backlog providing unusual forward visibility. My 12x is therefore reasonable and arguably conservative.
The implied equity value of approximately $2.66 trillion represents roughly 18% upside from the current $2.25 trillion market capitalization. That is not an extreme dislocation, but it is meaningful — particularly for a company of this quality, this scale, and this profitability trajectory.
Comparison to hyperscaler peers:
| Company | P/E | Revenue Growth | PEG | P/OCF |
|---|---|---|---|---|
| AMZN | 29x | 13.6% (consol) / 24% (AWS) | 0.97 | 16.1x |
| MSFT | ~33x | ~16% | ~1.5x | ~28x |
| GOOG | ~23x | ~14% | ~1.2x | ~20x |
| META | ~26x | ~22% | ~1.0x | ~18x |
Amazon is the cheapest on P/OCF and offers comparable or superior growth to Microsoft and Alphabet. The PEG discount to Microsoft (0.97 vs. ~1.5x) is notable — Amazon generates 5x the operating cash flow of Google Cloud alone yet trades at a meaningful discount to Alphabet.
I need to address the elephant in the room directly. The $200 billion FY26 capex guidance — roughly 28% of estimated FY26 revenue — is unprecedented in corporate history. It means TTM FCF will likely remain compressed in the $10-20 billion range through 2026, even as operating cash flow grows to perhaps $160-170 billion.
Two analysts on the call — Mark Mahaney and Doug Anmuth, both experienced Amazon observers — asked for financial guardrails or a minimum FCF floor. Management deflected both questions. The CFO pointed to "strong return on invested capital" without quantifying it. The CEO emphasized that capacity is "monetized as fast as we install it" and described it as an "extraordinarily unusual opportunity."
In my experience, management teams that refuse to provide financial guardrails during a major investment cycle fall into two categories: those who genuinely see demand so strong that constraining spend would be the wrong decision, and those who are empire-building without discipline. The $244 billion AWS backlog growing 40% and Trainium being "fully subscribed" suggest Amazon falls in the first category. But I would have preferred a clear statement — even something like "we expect to generate at least $20 billion in TTM FCF through the cycle."
This is the primary risk. If AI demand decelerates — whether from a macro downturn, a shift toward on-premise inference, or a technology inflection that reduces compute intensity — the return on $200+ billion in cumulative investment becomes questionable. I estimate AWS infrastructure depreciation will increase by $10-15 billion annually over the next two years, which will directly pressure AWS operating margins.
Amazon just reported a quarter that validated every element of the bull thesis: AWS re-acceleration, advertising durability, retail margin expansion, and the beginning of a custom silicon advantage. The forward-looking indicators — $244B backlog, Trainium fully subscribed, Bedrock up 60% QoQ — are as strong as any set of leading indicators I have seen from a mega-cap technology company.
The valuation at approximately 29x GAAP P/E (PEG ~1.0) is fair on a consolidated basis and meaningfully undervalued on a sum-of-parts basis. The operating cash flow multiple of 16x is genuinely cheap for a business generating $140 billion in annual operating cash flow growing at 20%.
The risk is concentrated in the capex cycle — 200billioninFY26withnocommittedFCFfloor.Butthedemandsignalsareunambiguous, thebalancesheetishealthy(57 billion net cash), and management has a track record of generating exceptional ROIC on prior infrastructure cycles (AWS's first decade of investment).
Recommendation: Amazon is not a position I would initiate in the high growth portfolio — it does not fit my concentration approach and the 12-14% consolidated growth rate is below my typical threshold. But for investors who hold it or are considering it, this quarter provides validation for a strong hold. The sum-of-parts discount (approximately 18% to estimated intrinsic value) creates a margin of safety. I would consider adding on any pullback toward the low $190s — which could easily occur if tariff fears, rate expectations, or a broader tech selloff creates an indiscriminate panic. And as I have observed many times, panics tend to be indiscriminate.
For those in my coverage universe focused on AI infrastructure, Amazon's results are also a strong read-through for Nebius, Pure Storage, and Arista — when the largest cloud provider is guiding $200 billion in capex and capacity is "monetized as fast as we install it," the infrastructure supply chain remains in a secular up-cycle.
Atlas scored AMZN at 4/5 conviction with a SOTP of ~3.2T.IagreewithAtlas′sdirectionalassessmentbutarriveatamoreconservativeSOTP( 2.66T, or ~18% upside vs. Atlas's ~45%). The difference is primarily in the AWS multiple — Atlas used 15x revenue versus my 12x. Given that AWS margins have compressed 450bps in three quarters and the $200B capex will add depreciation pressure, I think 12x is more prudent until we see margin stabilization. I also weight the FCF compression more heavily — Atlas acknowledged it as a risk but I view the lack of management commitment to a FCF floor as a more material concern for a company spending $200B per year.
Where Atlas and I align strongly: the AWS backlog divergence from revenue growth (40% vs. 24%) is a textbook bullish leading indicator, the PEG on earnings growth is genuinely attractive, and the Trainium custom silicon story is a structural margin advantage that the market is underappreciating.
Analysis by Bert Hochfeld. First analysis for AMZN — not a current portfolio holding. The company is not in the High Growth Portfolio or Income/Stability Portfolio.