[AMZN] Amazon: Q3 2026 earnings preview, can AWS growth outrun AI capex?

Amazon runs online retail, a third-party seller marketplace, advertising, Prime subscriptions and the cloud business AWS, and it manages them as three segments: North America, International and AWS[1]. In this Amazon Q3 2026 earnings preview, the company is scheduled to hold its call on 2026-10-29 to report the third quarter of 2026, ending September 30, 2026[2]. In the latest disclosed period, the second quarter of 2026, Amazon's net sales rose 20% to $200.6 billion and operating income reached $27.5 billion, up from $19.2 billion a year earlier; AWS revenue grew 37% to $42.2 billion and North America revenue grew 16% to $116.2 billion[3]. Net income was $62.6 billion, or $5.75 per diluted share, but it included $53.4 billion of pre-tax non-operating income, primarily from the investment in Anthropic[4]. Management guided third-quarter net sales to $197.0 billion to $202.0 billion, up 9% to 12%, said growth would be nearly 400 basis points higher excluding Prime Day in both 2025 and 2026, and built in an unfavorable currency impact of about 80 basis points; it guided operating income to $22.5 billion to $26.5 billion against $17.4 billion a year earlier, assuming no impact from energy derivative remeasurements[5]. The consensus compiled by Drillr from 27 analysts puts third-quarter revenue at $202.118 billion (range $200.107 billion to $204.409 billion) and EPS at $1.96 (range $1.82 to $2.17), so the revenue mean sits slightly above the top of company guidance[6]. The earnings-calendar figures are $202.441 billion of revenue and $2.01 of EPS; on the same basis, second-quarter EPS had been pegged at $1.83 against an actual $5.75[2], and almost all of that gap came from remeasurement gains on the Anthropic investment[4], so the third-quarter EPS comparison can be distorted by investment gains or losses in the same way.

Three things in this report matter most. The first is whether AWS can hold its growth and margin once one-off gains are stripped out: AWS growth excluding currency rose to 37% in the second quarter and its operating margin was 39.4%[7], but that profit included a roughly $600 million fair-value gain on energy derivative contracts[8], while AWS depreciation rose 67%, faster than revenue[9]; the third quarter is the first full quarter after the expanded OpenAI and Anthropic contracts, so it will test how quickly about $496 billion of unrecognized contract commitments turns into revenue[10]. The second is North America retail margin after Prime Day moved into the second quarter: the segment's operating margin was 7.9% in both the first and second quarters[7], but the second quarter included about $640 million of tariff refunds[11], and worldwide shipping costs grew 19%, faster than paid units at 17%[12]; the third quarter loses the event's absorption of fixed costs and still faces higher fuel and line-haul rates[13]. The third is whether the free cash flow gap keeps widening or starts to narrow: for the twelve months to June 30, operating cash flow was $161.4 billion, capital spending net of proceeds and incentives was $169.0 billion, and free cash flow turned into a $7.6 billion outflow[14]; the April shareholder letter set 2026 capital spending at about $200 billion[15], and a media account of the call said the plan had been raised to about $220 billion because of memory chip prices[16], so third-quarter cash flow and debt figures will show how much of this build the business can fund on its own.

Company Background and Business Structure

Amazon started as a retailer and is now a platform spanning e-commerce, advertising, subscriptions and cloud computing, led by President and CEO Andy Jassy[4]. Its annual report names seven customer sets it serves, namely consumers, sellers, developers, enterprises, content creators, advertisers and employees, and it evaluates and manages performance through the North America, International and AWS segments[1]. At the end of 2025 it employed about 1,576,000 full-time and part-time workers, and the business is seasonal, with the fourth quarter historically the highest-volume quarter[17].

Retail produces most of the revenue, while AWS produces most of the profit. Of 2025 net sales of $716.92 billion, North America contributed $426.31 billion, International $161.89 billion and AWS $128.73 billion; of $79.98 billion in operating income, AWS contributed $45.61 billion, North America $29.62 billion and International $4.75 billion[18]. By revenue type, 2025 online stores brought in $269.29 billion, physical stores $22.56 billion, third-party seller services $172.16 billion, advertising $68.64 billion and subscriptions $49.62 billion[19]. Amazon records its own product sales at the gross price, but in third-party sales it is not the seller of record and books only commissions plus fulfillment and shipping fees; advertising is recognized per click or impression, and subscriptions are mainly Prime fees and digital content[20]. In the second quarter of 2026, third-party sellers accounted for 61% of worldwide paid units[12].

In 2026 Amazon became both supplier and shareholder to the two leading AI labs. In the first quarter, OpenAI expanded its existing $38.0 billion multi-year commitment with AWS by $100.0 billion over 8 years; in the second quarter, Anthropic expanded its long-term commitment by more than $100.0 billion[10]. Over the same period, Amazon invested $28.7 billion in OpenAI's Series C preferred stock in the first half and funded the remaining $21.3 billion commitment after June 30, and it also invested $10.0 billion in Anthropic nonvoting preferred stock in the second quarter[21]. These private equity holdings are marked to observable price changes; the upward adjustment was $50.5 billion in the second quarter, primarily from the Anthropic preferred stock, and it flows through non-operating income rather than segment profit[22].

Financial History and Current Position

From 2023 to 2025, Amazon's revenue rose steadily while operating income more than doubled. Net sales went from $574.79 billion to $637.96 billion and then $716.92 billion; operating income rose from $36.85 billion to $68.59 billion and $79.98 billion, and net income was $30.43 billion, $59.25 billion and $77.67 billion[18]. Over the same years AWS operating income climbed from $24.63 billion to $45.61 billion, North America from $14.88 billion to $29.62 billion, and International swung from a $2.66 billion loss to a $4.75 billion profit[18]. 2025 operating income also absorbed a $2.5 billion FTC lawsuit settlement and $1.8 billion of the year's $2.7 billion in estimated severance costs, both recorded in the third quarter of 2025[23].

In the first half of 2026, both revenue growth and operating income accelerated. First-quarter net sales rose 17% to $181.5 billion (15% excluding currency), operating income was $23.9 billion and net income was $30.3 billion, including $16.8 billion of pre-tax gains from the Anthropic investment[24]. Second-quarter net sales were $200.61 billion and operating income was $27.46 billion, up 43%, for a 13.7% operating margin[14]; by segment, AWS earned $16.62 billion, North America $9.12 billion and International $1.72 billion[25], for operating margins of 39.4%, 7.9% and 4.1%[7]. Second-quarter net income of $62.65 billion far exceeded operating income because non-operating income for the quarter reached $53.40 billion[25].

On cash, operating cash flow is growing but capital spending is growing faster, and free cash flow has turned negative. In 2025, operating cash flow was $139.51 billion, capital spending net of proceeds and incentives rose from $77.66 billion in 2024 to $128.32 billion, and free cash flow fell from $38.22 billion to $11.19 billion[26]. For the twelve months ended June 30, 2026, operating cash flow was $161.40 billion, up 33%, capital spending was $169.01 billion, up 64%, and free cash flow was a $7.60 billion outflow, compared with an $18.18 billion inflow a year earlier[14]. On the balance sheet, cash and equivalents were $78.21 billion and marketable securities $44.78 billion at the end of June, while long-term debt rose from $65.65 billion at the end of 2025 to $128.89 billion[27]; Amazon then issued $25.0 billion of dollar notes in July[28] and sterling notes with an aggregate offering price of £4.242 billion in September[29].

Operating Model

Amazon's revenue runs on two different mechanisms, retail and cloud. Retail revenue is roughly paid units times average order value, but about 39% of units are Amazon's own products booked at the gross price and about 61% are third-party items that yield only commissions and logistics fees[12][20]; on top of the same shopping traffic sit advertising and subscription fees, which were $68.64 billion and $49.62 billion in 2025[19]. AWS revenue equals the compute, storage and AI services customers consume times price, and in the short run it is capped by capacity already online; its second-quarter annualized run rate was about $169 billion[8]. Unrecognized commitments carry a weighted-average remaining life of 6.4 years, and the timing of recognition depends on actual customer usage[10]. On seasonality, Prime Day moved from the third quarter into the second in most large markets in 2026, which the company said leaves third-quarter growth nearly 400 basis points below the trend excluding the event[30].

Operating profit is the sum of the three segments, but each segment makes money differently. The company treats operating income as a more meaningful measure than gross profit because its categories and services differ so widely[23]. AWS costs sit mainly in technology and infrastructure, which includes depreciation of servers, networking equipment and data centers along with rent and utilities[31][32], so its margin depends on the gap between revenue growth and depreciation growth: second-quarter AWS depreciation was $8.08 billion, up 67%[9], and since 2025 a subset of servers and networking gear has been depreciated over five years rather than six[33], so capital spending starts turning into depreciation roughly when equipment goes into service and continues for five to six years. The retail segments earn thinner margins, 6.9% for North America and 2.9% for International in 2025[7]; the company attributes their profit gains mainly to unit and advertising growth, partly offset by shipping, fulfillment and technology costs[34]. Shipping costs were $102.7 billion in 2025[35], and changes in fuel and freight rates reach North America profit largely within the same quarter.

Cash flow comes down to operating cash flow minus net capital spending. Retail collects from customers before paying suppliers, and accounts payable stood at $147.44 billion at the end of June, so this negative working capital supplies cash[27]; AI data center construction consumes cash now, while the matching depreciation returns to the income statement only over later years. Equity investments in OpenAI and Anthropic sit in investing activities, and acquisition and other investment payments reached $39.8 billion in the first half[21]; remeasurement gains on those stakes lift net income but bring in no cash[22]. When capital spending exceeds operating cash flow, the gap has to be filled from existing cash or new debt, and debt proceeds brought in $82.4 billion in the first half[21].

Industry and Competitive Position

In retail, Amazon competes in an extremely broad market against rivals from many industries, and it differentiates mainly through delivery speed and the Prime membership. The annual report calls its worldwide marketplace "evolving rapidly and intensely competitive," with rivals ranging from physical, online and omnichannel retailers to search engines, social networks and AI shopping tools, plus advertising and logistics providers[36]; specific rivals include omnichannel retailers such as Walmart, Costco and Target and cross-border platforms such as Temu and Shein. The shareholder letter says Amazon's grocery business exceeded $150 billion in gross sales in 2025, making it the second-largest grocer in the U.S.[15], and in the first half more than 40% more items reached Prime members same-day or overnight than a year earlier[37]. In advertising, on-site sponsored ads rely on shopping-intent data and compete with Google search ads and Meta social ads for brand and seller budgets.

In cloud, AWS is one of the largest public clouds and competes with Microsoft Azure, Google Cloud and Oracle, and in AI infrastructure it sells both Nvidia GPU capacity and its own Trainium chips. In the second quarter, both AWS's AI business and its chips business passed $25 billion annual revenue run rates, each growing at triple-digit percentages, and OpenAI and Anthropic both made multi-year, multi-gigawatt Trainium commitments[37]; management added that most capacity allocated for 2027 is already reserved and a significant share of 2028 capacity is pre-reserved[38]. The comparison has clear limits: Amazon does not disclose AWS AI revenue on its own, capacity size or a customer breakdown of the backlog, so an item-by-item comparison with peers is not possible. What sets Amazon apart from peers is that it is both a supplier and a shareholder of two AI labs, and that in 2026 it has used debt to cover the cash gap created by capital spending[27].

Core Debates

After AWS accelerated to 37% growth in the second quarter, can it hold growth above 35% and a margin above 37% in the third quarter without one-off gains?

AWS is both Amazon's main source of profit and the main destination of capital spending, so its growth and margin decide whether this build is expanding a high-margin business or piling up depreciation. In the second quarter AWS generated about 60% of operating income from about 21% of revenue, with $42.23 billion of revenue and $16.62 billion of operating income[25]. The case for continued strength is that AWS growth excluding currency stepped up from 17% in the second quarter of 2025 to 20%, 24%, 28% and then 37%[7]; Jassy called it five straight quarters of acceleration and confirmed that AWS AI power capacity is on track to double by the end of 2027 from 2025 levels[38]. Unrecognized commitments were about $496 billion at the end of the second quarter, up from about $244 billion at the end of 2025[10][39], and the first-quarter call had put the backlog at $364 billion[40].

The case against centers on profit quality and customer concentration. Second-quarter AWS profit included a roughly $600 million energy derivative gain[8], and without it the margin would be about 37.9%, while third-quarter guidance assumes no such remeasurement impact[5]. AWS depreciation of $8.08 billion equaled 19.1% of AWS revenue and grew 67%, far faster than revenue at 37%[9], and management acknowledged that higher prices for memory, SSDs and hard drives are raising AI infrastructure costs[41]. Most of the backlog comes from OpenAI and Anthropic, and Amazon holds equity in both[10]. An alternative reading is that the second-quarter jump came from a batch of capacity coming online at once and a staged ramp by large customers, so third-quarter growth could fall back toward 30% and the margin could slip as depreciation rises. The financial chain runs from new data center and Trainium capacity coming online, to customer consumption, to AWS revenue; in parallel, capital spending turns into depreciation inside technology and infrastructure costs, which sets the AWS operating margin.

What remains unresolved is how fast the backlog converts and what new capacity costs per unit. The observable third-quarter tests are whether AWS revenue reaches at least about $44.56 billion, or 35% growth; whether the AWS margin excluding one-off items stays at or above 37%; whether AWS depreciation in the 10-Q exceeds 21% of revenue; whether unrecognized commitments stay above $496 billion and include large new contracts beyond the two AI labs; and whether management reaffirms the goal of doubling AI power capacity by the end of 2027[30]. If AWS growth falls below 30% and the company attributes it to demand, or the margin excluding one-off items drops below 35%, the current view that capacity is turning into high-margin revenue would be falsified.

With Prime Day moved into the second quarter, can Amazon's North America retail business hold its third-quarter operating margin above 7% instead of losing it to fuel and freight inflation?

The North America margin directly decides which end of the guidance range third-quarter operating income lands on. North America contributes about 58% of revenue[25], so on third-quarter revenue above $110 billion, each percentage point of margin equals more than $1.1 billion of operating income, over a quarter of the $4 billion width of the guidance range[5]. The case for holding the margin is that North America earned 7.9% in both the first and second quarters, and worldwide paid unit growth rose from 15% in the first quarter to 17% in the second[7][12]; Grocery and Everyday Essentials grew meaningfully faster than the rest of the business[37], and same-day fresh grocery delivery for business customers now reaches more than 2,300 U.S. cities and towns[42]. Amazon also plans to more than double its fleet of Cardinal and Sparrow robotic arms in 2026 and, from April, added an FBA fuel and logistics surcharge that partly offsets higher freight rates[13].

The case against rests on transport costs and one-off items. Second-quarter worldwide shipping costs were $27.87 billion, up 19%, faster than the 17% rise in paid units[12][43]; management said Middle East conflict pushed up fuel and limited driver capacity pushed up line-haul rates, and only excluding those factors did shipping growth lag unit growth[13]. Second-quarter North America profit included about $640 million of tariff refunds, which the company described as the significant majority of refunds it expects[11], and excluding them the North America margin was about 7.3%. The third quarter has no Prime Day to spread fixed costs, and spending on the Amazon Leo satellite network is rising[32]. An alternative reading is that part of the first-half margin gain came from event timing and tariff refunds, and the third quarter will fall back below 7%. The financial chain runs from paid units, driven by Prime Day timing and grocery and essentials purchase frequency, to North America revenue; fuel and line-haul rates, regional inventory placement and order consolidation set shipping and fulfillment cost per unit, and therefore the North America operating margin.

The comparison base is itself noisy, and that is the main thing to keep in mind when reading the third quarter. The North America margin in the third quarter of 2025 was only 4.5% because the quarter carried the $2.5 billion FTC settlement[7][23], while the $1.8 billion of severance booked that quarter was not split by segment, so the year-over-year improvement will be overstated; worldwide paid units grew 11% in the third quarter of 2025, when Prime Day still fell in that quarter[12]. The observable tests are whether the North America margin stays at or above 7.3% and whether it still contains one-off items; whether worldwide shipping cost growth falls below paid unit growth; whether paid unit growth stays at or above 13%; whether North America revenue grows at least 11%; and what management says about fuel, line-haul rates and the FBA surcharge. If the North America margin drops below 6.9%, or shipping costs grow more than 3 percentage points faster than units, the current view that cost savings can absorb transport inflation would be falsified.

Advertising grew 26% in the second quarter; without Prime Day in the third quarter, can Amazon's ad business keep growing above 22%?

Advertising is the main source of Amazon's retail profit, and its growth sets how much cushion the retail margin has. Second-quarter advertising revenue was $19.81 billion, an annualized pace of about $80 billion[12], and the company's explanations for profit gains in both North America and International put unit and advertising growth first[44]. If ad growth drops below 20%, the retail margin would depend more on shipping and fulfillment savings. The case for continued strength is that advertising growth excluding currency held at 22% for four straight quarters from the second quarter of 2025 and rose to 26% in the second quarter of 2026[12]; the AI ad tool Ads Agent has expanded to 11 new countries this year, advertisers using it see 8% lower cost-per-impression and 6% lower cost-per-acquisition, active users of the AI shopping assistant Alexa for Shopping nearly doubled, and U.S. customers who use it spend over 40% more per order[42].

The case against is that the second-quarter acceleration coincided with Prime Day moving into the quarter. The third quarter is compared with the third quarter of 2025, which included Prime Day and produced $17.70 billion of advertising revenue[12]; second-quarter third-party seller services of $46.78 billion, up 16%, were also lifted by event timing. An alternative reading is that the second-quarter jump was mainly a timing effect, and third-quarter ad growth will fall back to around 20% or lower. The financial chain runs from shopping traffic and units to ad slots and video ad inventory, with Ads Agent lowering advertisers' acquisition costs, which together lift advertising revenue; third-party units generate commissions and FBA fees that form seller services revenue and ultimately feed North America and International segment profit.

What remains unresolved is how strong underlying ad demand is once event timing is removed, because the company does not report advertising profit separately. The observable tests are whether advertising grows at least 22%, or about $21.6 billion; whether third-party seller services grow at least 12%; whether the third-party share of units stays at or above 61%, against 62% in the third quarter of 2025[12]; and whether management cites incremental ad revenue from Prime Video ads, Ads Agent or the AI shopping assistant. If advertising growth falls below 18% or seller services growth falls below 9%, the current view that ad growth is resilient beyond event timing would be falsified.

Amazon's trailing twelve-month free cash flow has turned to an outflow of $7.6 billion; will the gap keep widening by about $10 billion a quarter in the third quarter, or start to narrow?

Whether operating cash flow can catch up with capital spending decides whether this AI build is funded by the business itself or keeps leaning on the debt market. Amazon planned roughly $200 billion of capital spending for 2026[15], doubled its long-term debt in the first half[27], and also committed tens of billions of dollars of equity funding to OpenAI and Anthropic[21]. Trailing twelve-month free cash flow went from $25.9 billion at the end of the first quarter of 2025 to $18.2 billion, $14.8 billion, $11.2 billion and $1.2 billion, and became a $7.6 billion outflow by the end of the second quarter of 2026, falling by roughly $9 billion to $10 billion in each of the last two quarters[14]. The case for a narrowing gap is that trailing operating cash flow growth has risen from 12% to 20% in 2025 to 33%, second-quarter operating income grew 43%[14], AWS revenue is accelerating, and capacity already online is producing revenue.

The case for a widening gap is just as concrete. Second-quarter cash capital spending was $53.1 billion, against $31.4 billion a year earlier, bringing the first half to $96.3 billion[21]; a media account of the call said the full-year plan was raised to about $220 billion because of memory prices[16], which implies about $62 billion a quarter in the second half, though that figure does not appear in the company's written filings. In July Amazon funded the remaining $21.3 billion of its OpenAI investment[21], and it issued debt in both July and September[28][29]. One reading holds that negative free cash flow is only the timing gap between construction and service, and that depreciation add-backs and new revenue will bring it back above zero within a year or two; the opposing reading is that capacity is consumed more slowly than it is built and the gap persists. The financial chain runs from AI data center and chip purchases to trailing net capital spending, and from operating income, depreciation add-backs and retail's negative working capital to trailing operating cash flow; the difference is free cash flow, which sets the scale of borrowing and the long-term debt balance.

What remains unresolved is when capital spending peaks and when cash returns from new capacity catch up. The observable third-quarter tests are whether trailing free cash flow stays above about -$12.6 billion or falls below -$20 billion; whether quarterly cash capital spending exceeds $62 billion and whether the full-year plan is raised again; whether trailing operating cash flow growth stays above 33%; and whether long-term debt at quarter-end exceeds about $160 billion. If free cash flow falls below -$20 billion or operating cash flow growth drops below 30%, the view that the gap is only a matter of timing would be materially weakened.

Risks and Falsifiers

The first risk is that investment remeasurement disconnects net income from operating performance. Amazon holds preferred stock in Anthropic and OpenAI whose carrying values move with observable prices and flow into net income; that produced $53.4 billion of pre-tax non-operating income in the second quarter, the bulk of the $62.6 billion in net income[4][22]. The consensus EPS of about $1.96 to $2.01 excludes such items, and a markdown would hit the income statement in the same way[6][2]. If third-quarter non-operating items return to within $2 billion and the gap between net income and operating income narrows, this risk would distort the reading much less.

The second risk is the overlap between customer and shareholder. The backlog grew from about $244 billion at the end of 2025 to about $496 billion at the end of June, and its weighted-average remaining life lengthened from 4.1 years to 6.4 years, with most of the increase coming from the expanded OpenAI and Anthropic commitments[39][10], while Amazon is also a major investor in both[21]. If either lab's funding or compute demand slows, utilization of the capacity already built and the pace of AWS revenue recognition would both suffer. If the backlog keeps growing in the third quarter and the company discloses new multi-year commitments from other enterprise customers, this concern would be weakened.

The third risk is depreciation outrunning revenue. Capital spending is turning into depreciation in bulk, new capacity may be consumed more slowly than it is built, and higher memory prices raise unit costs[41][9]. On second-quarter AWS revenue of $42.2 billion, each percentage point of margin lost removes about $420 million of quarterly operating income, and a rise in the depreciation ratio from 19.1% to 21% equals about $800 million. If the third-quarter AWS margin excluding one-off items stays at or above 37% and the depreciation ratio stays at or below 20%, this risk has not materialized.

The fourth risk is transport inflation. Middle East conflict is lifting fuel prices and limited driver capacity is lifting line-haul rates[13], while the third quarter lacks a major sales event to spread costs. Worldwide shipping costs were $27.87 billion in the second quarter, and each percentage point by which shipping growth exceeds unit growth adds about $280 million of quarterly cost, falling mainly on North America operating income[12]. If third-quarter shipping growth is no higher than unit growth and the North America margin stays at or above 7.3%, this risk is falsified.

The fifth risk is an advertising slowdown. Part of the second-quarter acceleration came from event timing, and the third quarter is measured against a base that includes Prime Day, which would magnify any slowdown in underlying demand[5]. Against third-quarter 2025 advertising revenue of $17.7 billion, every 4 percentage points of lost growth removes about $700 million of high-margin revenue, mainly affecting North America and International operating income[12]. If third-quarter advertising growth stays at or above 22%, this risk has not materialized.

The sixth risk is that capital spending keeps exceeding operating cash flow, the gap depends on borrowing, and interest expense rises. Second-quarter interest expense was $1.314 billion, against $516 million a year earlier[14]; the $25.0 billion of notes issued in July carry effective interest rates between 4.67% and 6.33%, and at a rough 5% that is about $1.25 billion of annual interest[28]. If trailing free cash flow falls by no more than $5 billion in the third quarter and the company announces no new debt issue, this risk would ease for now.

What to Watch Next

• AWS capacity turning into revenue and profit: second-quarter growth was 37% and the margin 39.4%, including about $600 million of energy derivative gains. Watch whether revenue reaches about $44.56 billion and whether the margin excluding one-off items holds at 37% or more; growth below 30% attributed to demand, or a margin below 35%, would falsify the current view.
• AWS depreciation and backlog: depreciation was 19.1% of AWS revenue and the backlog about $496 billion. Watch whether the ratio exceeds 21% and whether the backlog keeps growing with large contracts beyond the two labs; growth plus other enterprise commitments would strengthen the view.
• North America retail margin: 7.9% in the second quarter (about 7.3% excluding tariff refunds) against 4.5% in the third quarter of 2025. Watch whether it holds at 7.3% or more and whether one-off items remain; a margin below 6.9% would falsify the view.
• Shipping versus paid units: second-quarter shipping costs rose 19% against 17% unit growth. Watch whether shipping growth falls below unit growth and whether unit growth holds at 13% or more; shipping running more than 3 points ahead of units would falsify the view.
• Advertising and seller services: advertising grew 26% and seller services 16%, against a third-quarter 2025 ad base of $17.7 billion. Watch whether ads grow at least 22% (about $21.6 billion) and seller services at least 12%; ads below 18% or seller services below 9% would falsify the view.
• Free cash flow and capital spending: trailing free cash flow was -$7.6 billion and second-quarter capital spending $53.1 billion. Watch whether free cash flow stays above about -$12.6 billion and whether quarterly spending exceeds $62 billion; free cash flow below -$20 billion would falsify the timing-gap reading.
• Operating cash flow and debt: trailing operating cash flow grew 33% and long-term debt stood at $128.89 billion. Watch whether growth holds above 33% and whether debt exceeds about $160 billion; growth below 30% would weaken the view.

Conclusion

Amazon runs on two engines: retail earns thin but improving margins from units, advertising and membership fees, while AWS earns most of the profit from how fast its capacity is consumed. Second-quarter operating income was $27.5 billion, up 43%, with AWS contributing $16.6 billion[3]; yet trailing twelve-month free cash flow through June had turned into a $7.6 billion outflow, and long-term debt doubled within six months to $128.9 billion[14][27]. The central unresolved relationship is whether about $496 billion of contract commitments and 37% AWS growth can turn capital spending on the order of $200 billion into revenue and operating cash flow fast enough, rather than into depreciation and debt first[10].

Two independent assessments published after the second-quarter report offer complementary readings with different emphasis. Keithen Drury of The Motley Fool argues that a 39% AWS margin, growth accelerating to 37%, a backlog of nearly $500 billion and AWS supplying about 60% of operating income show rising margins and plenty of work ahead, which in his view justifies putting all of Amazon's cash flow into AI data centers and spending about $220 billion this year[45]; this is the optimistic side of the AWS debate, and it assumes the backlog converts on schedule and depreciation does not compress margins. Rimmi Singhi of Zacks argues instead that the core business is converting profit into cash just as well, and that the problem appears one step further down the cash flow statement: revenue benefits from capacity already in use while free cash flow absorbs the cost of capacity still being built, and the pressure should ease only as new data centers come online and generate revenue[46]. Both treat negative free cash flow as a cost of the build phase rather than a return problem; they differ in emphasis on timing and risk, with the first reading high margins as evidence the spending is justified and the second placing relief explicitly after new capacity enters service. These are outside interpretations, not facts, and they do not amount to a vote.

What would change the assessment is a combination of later operating and financial data. If AWS holds growth above 35% and a margin above 37% in the third quarter after removing the energy derivative gain, the North America margin stays near 7.3% without tariff refunds or Prime Day, trailing operating cash flow keeps growing above 33% and the decline in free cash flow narrows, the view that the spending is turning into returns would be clearly strengthened. Conversely, if AWS growth slips toward 30% while the depreciation ratio rises past 21%, North America shipping costs keep outrunning units, free cash flow falls below -$20 billion and long-term debt keeps climbing quickly, this build would look more like widening a cash gap than expanding a high-margin business.

Sources

[1] AMZN 10-K filed 2026-02-06 · business overview and customer sets · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[2] Drillr earnings calendar (updated 2026-09-28) · AMZN 2026-10-29 call · 2026-09-28 · Drillr earnings calendar · https://gateway.drillr.ai/mcp/private

[3] AMZN 8-K filed 2026-07-30 · Q2 2026 results · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[4] AMZN 8-K filed 2026-07-30 · Q2 2026 net income and cash flow · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[5] AMZN 8-K filed 2026-07-30 · Q3 2026 guidance · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[6] Drillr analyst_financial_estimates (updated 2026-09-28) · AMZN quarter ending 2026-09-30 · 2026-09-28 · Drillr analyst_financial_estimates · https://gateway.drillr.ai/mcp/private

[7] AMZN 8-K filed 2026-07-30 · supplemental segment metrics Q1 2025-Q2 2026 · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[8] AMZN Q2 2026 earnings call 2026-07-30 · segment performance · 2026-07-30 · earnings-call · https://gateway.drillr.ai/mcp/private

[9] AMZN 10-Q filed 2026-07-31 · depreciation by segment Q2 2026 · 2026-07-31 · 10-Q · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm

[10] AMZN 10-Q filed 2026-07-31 · AWS commitments and OpenAI/Anthropic expansions · 2026-07-31 · 10-Q · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm

[11] AMZN 10-Q filed 2026-07-31 · IEEPA tariff refunds · 2026-07-31 · 10-Q · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm

[12] AMZN 8-K filed 2026-07-30 · supplemental net sales mix and unit metrics · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[13] AMZN Q2 2026 earnings call 2026-07-30 · fulfillment costs and transportation · 2026-07-30 · earnings-call · https://gateway.drillr.ai/mcp/private

[14] AMZN 8-K filed 2026-07-30 · supplemental TTM cash flow and results · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[15] AMZN 8-K filed 2026-04-08 · 2025 shareholder letter · 2026-04-08 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000110465926041034/

[16] Yahoo Finance (Benzinga) 2026-08-01 · Amazon Raised 2026 CapEx by $20 Billion · 2026-08-01 · Benzinga(Yahoo Finance 转载) · https://finance.yahoo.com/markets/stocks/articles/amazon-raised-2026-capex-20-223016275.html

[17] AMZN 10-K filed 2026-02-06 · seasonality and employees · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[18] AMZN 10-K filed 2026-02-06 · segment results FY2023-FY2025 · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[19] AMZN 10-K filed 2026-02-06 · net sales by product and service FY2025 · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[20] AMZN 10-K filed 2026-02-06 · revenue recognition by activity · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[21] AMZN 10-Q filed 2026-07-31 · Q2 2026 cash capex and investments · 2026-07-31 · 10-Q · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm

[22] AMZN 10-Q filed 2026-07-31 · Anthropic valuation gains · 2026-07-31 · 10-Q · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm

[23] AMZN 10-K filed 2026-02-06 · 2025 operating income and Q3 2025 charges · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[24] AMZN 8-K filed 2026-04-29 · Q1 2026 results · 2026-04-29 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000012/amzn-20260331xex991.htm

[25] AMZN 8-K filed 2026-07-30 · Q2 2026 segment information · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[26] AMZN 10-K filed 2026-02-06 · free cash flow reconciliation FY2025 · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[27] AMZN 8-K filed 2026-07-30 · balance sheet June 30 2026 · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[28] AMZN 10-Q filed 2026-07-31 · July 2026 notes issuance · 2026-07-31 · 10-Q · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm

[29] AMZN 8-K filed 2026-09-14 · sterling notes offering · 2026-09-14 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000110465926107526/tm2624614d5_8k.htm

[30] AMZN Q2 2026 earnings call 2026-07-30 · Q3 guidance and capacity · 2026-07-30 · earnings-call · https://gateway.drillr.ai/mcp/private

[31] AMZN 10-K filed 2026-02-06 · fulfillment and technology cost definitions · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[32] AMZN 10-K filed 2026-02-06 · infrastructure cost definition · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[33] AMZN 10-K filed 2026-02-06 · server useful lives · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[34] AMZN 10-K filed 2026-02-06 · 2025 segment operating income drivers · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[35] AMZN 10-K filed 2026-02-06 · shipping costs FY2025 · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[36] AMZN 10-K filed 2026-02-06 · competition · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[37] AMZN 8-K filed 2026-07-30 · Q2 2026 CEO remarks and AI run rates · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[38] AMZN Q2 2026 earnings call 2026-07-30 · AWS capacity and margin Q&A · 2026-07-30 · earnings-call · https://gateway.drillr.ai/mcp/private

[39] AMZN 10-K filed 2026-02-06 · AWS commitments not yet recognized December 2025 · 2026-02-06 · 10-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm

[40] AMZN Q1 2026 earnings call 2026-04-29 · AWS backlog and Rufus · 2026-04-29 · earnings-call · https://gateway.drillr.ai/mcp/private

[41] AMZN Q2 2026 earnings call 2026-07-30 · stated risks · 2026-07-30 · earnings-call · https://gateway.drillr.ai/mcp/private

[42] AMZN 8-K filed 2026-07-30 · Ads Agent and Alexa for Shopping · 2026-07-30 · 8-K · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000024/amzn-20260630xex991.htm

[43] AMZN 10-Q filed 2026-07-31 · Q2 2026 shipping costs · 2026-07-31 · 10-Q · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm

[44] AMZN 10-Q filed 2026-07-31 · Q2 2026 segment operating income drivers · 2026-07-31 · 10-Q · https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm

[45] The Motley Fool 2026-09-23 · Amazon's AWS Backlog Climbed to $496 Billion as Its Cloud Margin Reached 39% · 2026-09-23 · The Motley Fool · https://www.fool.com/investing/2026/09/23/amazons-aws-backlog-climbed-to-496-billion-as-clou/

[46] Zacks via Yahoo Finance 2026-09-16 · Amazon's AI Bet Is Paying Off in Revenues: Why Isn't FCF Following? · 2026-09-16 · Zacks Investment Research(Yahoo Finance 转载) · https://finance.yahoo.com/technology/ai/articles/amazons-ai-bet-paying-off-145400682.html