Our prior take argued the market was finally pricing the AWS capex acceleration as an investment rather than a drag. Six months later, capex has stepped again and FCF has compressed sharply. The thesis needs a meaningful update.
Our previous Amazon deep dive (Amazon's Margin Inflection: The Capex Bet the Market Is Finally Taking Seriously) argued that the elevated AWS capex cycle was finally being priced by the market as an investment rather than a drag. The thesis rested on operating income compounding faster than free cash flow during the investment phase, and a re-rating back to normalised FCF multiples once the capex moderated.
The FY25 print has landed and the capex story has actually gotten larger, not smaller. Amazon's FY25 capex hit $131.8 billion, up from $83.0 billion in FY24. Free cash flow compressed from $32.9 billion in FY24 to $7.7 billion in FY25, a 77% decline. Operating income continued to compound, reaching $80.0 billion against $68.6 billion the prior year.
This is an Update piece. The question: does the larger capex number break the thesis or extend it? The Research Desk's answer is that it extends it, but with a meaningful downward revision to the 12-month fair value. The operating income trajectory is intact and the AWS revenue growth rate has held. The timeline for the FCF recovery has shifted out by approximately 12 months, which reduces the expected return over our original time horizon.
Amazon trades at $2.69 trillion of market cap, a trailing P/E of 35 and a forward P/E of 31. The prior fair value of $290 has been revised to $265, reflecting the extended capex window and the compounded FCF compression. That still implies 8-10% upside from the current price near $225. The thesis holds at reduced conviction, not at the prior high conviction.
One tactical observation worth including in the update. The stock's current price of approximately $225 compared to the 200-day moving average of $226 puts the equity almost exactly at the long-term trend line. That is the kind of level that has historically been a durable accumulation zone for long-term holders; above this level, patience is the move, and below this level, incremental accumulation has produced favourable outcomes.
Three primary changes since the prior take. First, AWS capex accelerated materially in H2 FY25 as demand for AI compute capacity continued to outpace supply. Management has guided that capex intensity remains elevated through FY26 and likely into H1 FY27.
Second, operating income growth exceeded expectations. The FY25 step from $68.6 billion to $80.0 billion is approximately 17% growth, above the 14-15% the prior thesis modelled. Advertising, AWS, and retail margin accretion all contributed.
Third, the retail margin continued its structural improvement. The North America segment operating margin crossed 6% on a run-rate basis in FY25, up from approximately 4.5% in FY24. The international segment reached positive margin for the first time in four years. Retail contribution to consolidated operating income has accelerated even as the capex headline dominates the commentary.
The combination of these three changes supports a revised view: the short-term cash flow compression is worse than modelled, but the medium-term operating income trajectory is better. The thesis rebalances around a longer duration for the return, compensated by a stronger normalised operating income base.
TickerXray Report
Get the complete Amazon report with all 12 quantitative models, AI-generated investment thesis, and real-time data.
The FCF compression from $32.9 billion to $7.7 billion is a $25 billion decline in a year when operating income expanded by $11 billion. The entire $36 billion gap is accounted for by the capex step (up $49 billion) minus depreciation increase (up approximately $12 billion). Working capital moves and interest expense absorb the residual.
The question is when the capex-to-operating-cash-flow ratio normalises. In a normalised regime, capex runs at approximately 6-8% of revenue (retail) plus AWS-specific infrastructure investment. At $720 billion of FY26 revenue, the retail capex baseline should run at $45-55 billion. AWS infrastructure capex, at a normalised pace supporting 22-25% revenue growth, should run at $55-65 billion. Total normalised capex would therefore be $100-120 billion annually.
FY26 capex consensus sits at approximately $140-150 billion, above the normalised range. FY27 consensus sits at approximately $130-140 billion, still above. FY28 consensus runs at approximately $110-125 billion, entering the normalised range.
On that timeline, FCF recovers slowly. FY26 FCF is likely to sit around $5-15 billion, FY27 around $25-40 billion, and FY28 around $55-75 billion. At $70 billion of normalised FCF on a $2.7 trillion market cap, the yield is 2.6%, consistent with comparable mega-cap compounders. The multiple re-rates once investors see FY27 numbers confirm the FCF recovery trajectory, which is approximately 18-24 months from today.
The Research Desk's prior thesis assumed this timeline was 12 months. The revision stretches it to 18-24 months. That extension is the reason for the fair value revision from $290 to $265.
One further refinement on the normalised FCF math. The working-capital dynamic embedded in the retail business naturally runs as a source of cash during Q4 (seasonal inventory release) and a use of cash during Q2-Q3 (inventory build). That seasonality distorts quarterly FCF prints but averages out on a full-year basis. The FY25 FCF number is therefore a reasonable read on the run-rate, not a quarterly anomaly.
Rebuild the fair value calculation on the updated trajectory. The prior thesis valued Amazon at $290 per share on a sum-of-the-parts: AWS at 20x forward EBITDA, retail at 12x, advertising at 25x, and a holding-company modest discount. The prior AWS EBITDA multiple was based on a comparison to Microsoft Azure peers at similar growth rates and margin profiles.
The updated SOTP: AWS forward EBITDA for FY26 is approximately $85 billion (vs prior $78 billion) on a 24% segment growth rate. At 19x forward EBITDA (modestly compressed versus prior 20x to reflect macro multiple compression), AWS is worth $1.62 trillion. Retail forward operating income for FY26 is approximately $22 billion (vs prior $18 billion). At 14x, retail is worth $308 billion. Advertising revenue of $70-75 billion at a 32% segment operating margin gives $22 billion of EBIT; at 22x, advertising is worth $485 billion.
Add the three and the sum is $2.41 trillion. Adjust for net cash of approximately $86 billion and subtract $200 billion of minority value-transfer friction (consistent with large-cap conglomerate structures), the implied equity value is approximately $2.30 trillion, or $220 per share. Add the optionality value of the next four-to-five years of capex-driven revenue compounding, valued at 20% of base-case equity value, and the total is approximately $2.76 trillion, or $265 per share.
The bear case drops the AWS growth rate to 18% in FY27 and multiple to 16x, takes retail EBIT to $18 billion at 12x, and trims advertising by 5% of multiple. That produces a $190 per share bear value. The bull case accelerates AWS revenue growth, re-rates AWS to 22x EBITDA, and lifts retail margin a further 100 basis points. That produces a $310 per share bull value.
The Research Desk's base case target is $265 with a bear floor of $190 and bull stretch at $310. Current price of $225 is modestly below the base case but well above the bear case. Risk-reward is positively skewed but less attractive than the prior thesis implied.
The operating income line is the most important diagnostic for the updated thesis. FY25 operating income of $80 billion on $717 billion of revenue gives an 11.2% operating margin. That is the highest consolidated operating margin Amazon has ever produced. The margin path from 2.4% in 2022 to 11.2% in 2025 is a structural improvement, not a cyclical one.
Retail contributed approximately 4-5 percentage points of consolidated margin accretion over that window, through efficiency improvements, advertising attachment, and pricing discipline. AWS contributed 2-3 percentage points through infrastructure efficiency and AI services mix. Advertising contributed 1-2 percentage points as the margin mix shifted toward high-attachment digital placements.
The forward operating income trajectory depends on each of those levers continuing to compound. Retail has further margin runway, probably another 100-200 basis points over 2-3 years. AWS has limited incremental margin expansion but revenue compounding at 24% produces operating income growth of 25-30% annually. Advertising continues to accrete at the 15-20% annual pace it has been running.
Stack them together and FY26 operating income lands at $94-100 billion, FY27 at $110-120 billion, FY28 at $128-140 billion. Even with the FCF compression, the operating income compounding supports the core thesis. The equity return gets delivered through earnings growth plus multiple mean-reversion once the FCF normalises. The timing is the variable; the direction is not.
Historically, capex-driven compression cycles at compounders have produced total returns of 12-18% annually through the investment phase (compounding earnings against a compressed multiple) and 25-35% annually in the 18-24 months following capex peak (as the multiple re-rates on recovering FCF). Amazon's set-up fits the pattern. The updated thesis captures both phases but weights the first phase more heavily given the extended capex trajectory.
The risks have shifted in composition. First, AWS demand deceleration. If the AI compute demand that justified the capex acceleration slows before the capacity comes online, Amazon ends up with substantial unutilised infrastructure. The consequence would be depreciation absorbing margin with minimal incremental revenue to offset. Probability: 15-25% on the 24-month window.
Second, retail margin reversal. The retail margin expansion has been steady but not yet proven through a recession. A meaningful US consumer weakness could compress retail margin back toward the 3-4% range. The FY26 and FY27 macro outlook will determine whether this risk materialises.
Third, competitive pressure from hyperscaler peers on AWS pricing. Both Azure and Google Cloud have been competing aggressively for enterprise AI workloads. Margin compression on the AWS segment would reduce the forward earnings power.
The Research Desk rates the aggregate risk profile as modestly higher than the prior update implied, reflecting primarily the capex timeline extension and the broader macro uncertainty. Position sizing should be more conservative than the prior thesis suggested.
Amazon, Microsoft and Alphabet are all in elevated capex phases simultaneously. The composition of each capex envelope differs in instructive ways. Microsoft's $64.6 billion of FY25 capex represents 23% of revenue, with most of the spend going to AI infrastructure. Alphabet's $75 billion of 2025 capex is approximately 21% of revenue. Amazon's $131.8 billion is 18% of revenue but the absolute number is the largest of the three.
The relative capex intensity tells investors that Amazon's capex commitment is somewhat lower as a percentage of revenue than its hyperscaler peers, despite the larger absolute number. That is consistent with Amazon's revenue base being heavier in retail (which is less capex-intensive than cloud infrastructure). The composition matters for the FCF recovery shape: Amazon's path back to normalised FCF is somewhat smoother because the retail segment continues to generate cash while the AWS segment absorbs investment.
The forward FCF yield for the three is also instructive. Microsoft's normalised FCF yield is approximately 2.5% based on a reasonable FY27 trajectory. Alphabet's is approximately 3.0%. Amazon's, on the FY27 recovery path, is approximately 2.6%. The yields converge once the capex moderates, suggesting the long-term valuation profiles for the three are similar.
The distinction is timing. Amazon's capex cycle peaks later than Microsoft's; the FCF normalisation is correspondingly later. For investors choosing between the three on a 12-18 month basis, Microsoft has a modestly earlier FCF recovery profile and is therefore the more conservative position. For investors with a 24-month-plus horizon, Amazon's larger absolute capital commitment translates to a larger absolute revenue base when the cycle resolves, which produces commensurately larger absolute earnings power.
The Amazon thesis extends rather than breaks on the larger capex numbers. The fair value has been revised down from $290 to $265, reflecting the extended timeline to FCF recovery. The core compounding machinery (AWS growth, retail margin expansion, advertising mix) remains intact and in some dimensions has exceeded the prior base case.
The trade is to hold existing positions, accumulate on any pullback below $210, and set a 12-month target of $265 with a bear-case floor review at $190. The catalyst calendar runs: Q1 FY26 print in late April, the AWS segment capacity and utilisation updates, the advertising revenue disclosures, and the FY27 preliminary capex outlook. Two of those clearing in favour of the thesis supports the re-rate toward fair value. The Research Desk rates this a hold with conviction, moderately reduced from the prior Update call. The stock is not a high-conviction buy at current prices but it remains a high-quality core holding for long-term compounders.
Full forensic analysis of Amazon
+ 6 more models included
150,000+ stocks covered
Global coverage across 60+ exchanges. Every report includes all 12 quantitative models and AI analysis.
View plansEvery report runs 12 quantitative models and generates an AI investment thesis. From Piotroski scores to manipulation detection -- get the full picture in seconds.
12 forensic models
Piotroski, Altman, Beneish, DuPont & more
AI investment thesis
Synthesized outlook on every stock
Manipulation detection
Spot red flags before they hit the news
150,000+ tickers
Global coverage across 60+ exchanges
Expected return
Forward return projections for every stock
Real-time data
Live prices, insider trades, news sentiment
Free accounts get 1 report per month. Pro gets unlimited.
AMZN's FY2025 revenue hit $716.9B with net income of $77.7B. AWS, advertising, and the maturing retail margin form a three-engine compounding profile that the FCF compression is hiding.
Advertising revenue has scaled into a three-digit billions business with margins far above the retail complex. The Signals Desk reads the segment as more important than the consensus models price it.
The BofA and KeyBanc upgrades to $325 capture most of the thesis. The Valuation Desk thinks fair value extends higher once capex normalises.