Amazon At $252 Is A Fair Value Question With An Easy Answer

Amazon's capex grew from $52.7 billion in FY2023 to $131.8 billion in FY2025, a 150% increase over two years. The absolute dollar magnitude is staggering. The purpose of that capex is AI infrastructure to support Bedrock, Trainium, and the broader AWS AI workload expansion. The capex question decomposes into two sub-questions. First, will the AI infrastructure investment generate returns that exceed cost of capital. Second, when does the capex cadence normalise to a steady-state run rate.

On the first question, AWS AI service revenue is currently running at approximately $12-15 billion annualised and growing at 100%-plus year over year. That revenue carries gross margins in the 55-65% range given the specialised hardware and software stack. Applied to the capex base, the payback on the AI investment runs at roughly 3-4 years on current run rates. That is an unambiguously positive return profile by cloud infrastructure standards.

On the second question, management has signalled in the most recent earnings call that FY2026 capex will be elevated but not growing at the same pace as FY2025. The implied trajectory is capex in the $125-140 billion range in FY2026, normalising to $85-95 billion in FY2027 as the initial AI infrastructure buildout cycles complete. That normalisation is what unlocks the free cash flow thesis.

AWS revenue hit approximately $123 billion in FY2025 and is growing at 21% year over year as of the most recent quarter. That growth rate is the highest AWS has posted in two years. The acceleration is being driven by two specific dynamics: enterprise AI workloads moving from experimentation to production, and previously-paused cloud migrations resuming under the new AI infrastructure layer.

AWS operating margin sits at 37.5% on a consolidated basis, up from 29.9% in FY2023. That margin recovery has occurred despite the heavy capex cycle, because the revenue growth rate has outpaced the infrastructure expense growth. At scale, cloud infrastructure has operating leverage that bend up as the revenue base expands. AWS is in the favourable part of that leverage curve.

Relative to Microsoft Azure and Google Cloud, AWS retains the largest absolute market share but has been losing share in the AI-specific workload categories to Microsoft. The BofA upgrade explicitly flagged share stabilisation in these categories as the specific driver of the target increase. The Valuation Desk agrees; Trainium adoption, Bedrock expansion, and the Anthropic partnership have each materially strengthened AWS's AI-native workload position over the past twelve months.

North American retail operating margin reached 6.4% in FY2025 from 4.1% in FY2023. International retail turned positive at 2.8% from minus 3.2% over the same period. The combined retail business is now generating approximately $27 billion of operating profit against the $11 billion loss posted in FY2022. The swing is worth roughly $38 billion of enterprise value on its own at peer retail multiples.

The drivers of the retail margin expansion are well-understood and durable: fulfilment network restructuring, advertising monetisation of the retail flywheel, and Prime subscription economics. All three are continuing tailwinds rather than one-time benefits. The Valuation Desk models North American retail margin expanding to 8-9% by FY2027 on continued advertising attachment and fulfilment optimisation.

At 8% North American retail margin on roughly $440 billion of segment revenue by FY2027, that produces $35 billion of operating profit from North American retail alone. Compare to Walmart's U.S. segment producing roughly $23 billion of operating profit on $480 billion of revenue, and Amazon's retail business is delivering structurally higher margin than any peer. Advertising is the reason.

Amazon advertising revenue hit approximately $54 billion in FY2025 and is growing at 19% year over year. Gross margin on advertising is estimated above 70%. Operating margin after infrastructure and personnel allocation is estimated above 50%. This segment alone is generating $27-30 billion of operating profit. Applied to peer comparable multiples (24-28x forward earnings for dedicated ad-tech businesses), the segment is worth $700-850 billion of enterprise value.

The market is currently ascribing effectively zero separate value to the advertising business within the Amazon consolidated valuation. It is simply absorbed into the aggregate operating profit line. Separated and valued on merit, the advertising segment alone represents 26-31% of Amazon's market cap. That is before assigning any value to AWS, retail, Prime, or the logistics network.

The analytical implication is straightforward. If the market ever begins valuing Amazon on a sum-of-parts basis rather than a consolidated multiple basis, the fair value moves materially higher than the current $252. Historical precedent for this revaluation is Alphabet in 2017-2018 when the market began separately valuing YouTube, Cloud, and Search in sell-side models. The Alphabet multiple expanded by roughly 30% over the subsequent eighteen months on exactly that narrative shift.

Prime subscription revenue hit approximately $47 billion in FY2025 and is growing in the high single digits. The Prime ecosystem, which combines subscription fees, advertising within Prime Video, and the retail flywheel contribution, is worth considering as a separate unit economic analysis. The combined Prime economics produce what the Valuation Desk estimates at 18-22% operating margin, or roughly $8-10 billion of operating profit attributable to the subscription franchise before ancillary benefits.

The ancillary benefits are larger than the subscription segment itself. Prime members purchase 2.5x the frequency of non-members on Amazon retail, spend 3-4x more in annual retail GMV, and are 80% more likely to convert advertising impressions. Those behavioural metrics mean that Prime is simultaneously a subscription business and the most valuable customer segmentation tool in global retail. Neither Walmart+ nor Target Circle 360 has achieved comparable engagement characteristics.

Prime Video advertising, in particular, has not yet reached steady-state monetisation. The Valuation Desk estimates Prime Video ad revenue at $3.5-4.5 billion annualised today, with a 24-month path to $10-12 billion as the ad supply and demand ecosystem mature. That additional $7-8 billion of high-margin revenue is not in consensus models. It will be.

Across three mega-cap capex cycles, the pattern has been consistent. AWS itself is the cleanest precedent. Between 2014 and 2017, AWS invested aggressively ahead of demand. Free cash flow compressed. Analysts questioned the return profile. The infrastructure matured into the 2018-2020 harvest cycle, and AWS operating profit grew from $4.3 billion in FY2017 to $13.5 billion in FY2020. The stock compounded at 25% annually across that window.

The current AI infrastructure cycle has the same shape. Capex is heavy now. Revenue conversion is building. Free cash flow compression is the temporary cost of the investment cycle. Historically, the market has compressed the multiple during the investment phase and expanded it during the harvest phase. Amazon is currently in the investment phase. The harvest phase is twelve to eighteen months away on the Valuation Desk's timeline.

The last time Amazon traded at a clear dislocation of this type was Q1 2022, with the stock at $100 equivalent and AWS operating income temporarily compressed on the FX and power cost increases of that specific quarter. The subsequent three-year return was approximately 175%. The setup today is not identical but has similar structural features: a visible temporary negative, a clear underlying trajectory, and a market audience slow to price the resolution.

Amazon reports Q1 earnings in the coming weeks. The Valuation Desk will be watching three specific data points. First, AWS revenue growth rate; consensus is 19-20%, and the desk's estimate is 21-22%. A beat on this metric is the cleanest read on the AI inference monetisation thesis. Second, North American retail operating margin; consensus is 6.6%, and the desk's estimate is 7.0-7.2%. A beat here strengthens the retail reset thesis. Third, capital expenditure commentary for FY2026; consensus is $135 billion, and the desk's estimate is $125-140 billion. Any commentary that suggests an earlier-than-expected moderation would be a meaningful catalyst.

Any two of these three data points delivering above consensus would confirm the BofA and KeyBanc upgrade path to $325. All three delivering above consensus would accelerate the re-rating. If all three underwhelm, the capex fatigue debate intensifies and the stock could compress toward $220-230 in the near term. The probability distribution across these three scenarios is roughly 35%, 35%, and 30% in the Valuation Desk's framework.

At the consolidated level, Amazon trades at 30.9x forward earnings. Microsoft trades at 22x. Alphabet trades at 29.9x. Amazon's premium to Microsoft is arguably justified by the retail operating leverage still to come and the advertising segment acceleration. The premium to Alphabet is narrower and reflects similar AI capex dynamics.

On enterprise value to sales, Amazon trades at 3.7x forward revenue. The cloud-pure-play peer set (Microsoft cloud, Snowflake, Datadog) trades at 7-12x revenue. Amazon's lower multiple reflects the lower-margin retail business dragging the aggregate down. Strip out retail and value the remaining cloud-plus-ads business at 8x revenue, and the residual retail business is being valued at less than 1x revenue. That is structurally wrong for a business generating $27 billion of operating profit with expanding margins.

The argument is ultimately simple. Amazon's valuation is mispriced because the consolidated structure obscures the underlying economics of the profitable high-growth segments. The BofA and KeyBanc upgrades are a small step toward the sum-of-parts reframing that this business needs. More analysts will follow.