# Amazon: retail scale, cloud profits, and the cash cost of expansion

![AWS profits, retail scale, heavy reinvestment; original conceptual business diagram.](https://stuff.ac/api/v1/items/466c0495-41f0-463a-b565-1a866b7fcc89/content)

*Original schematic explaining the business. Shapes are conceptual and are not product photographs or proportional money flows.*

**Research date:** October 6, 2026 · **Ticker:** NASDAQ: AMZN · **Coverage:** 5 of 10  
**Latest reported quarter:** Q2 2026, ended June 30; published July 30, 2026.  
**Currency:** USD billions unless stated. Market value uses the space's October 6 premarket snapshot. Amazon reports on a calendar year.

## Quick overview

Amazon combines a large retail and marketplace network with advertising, subscriptions, logistics services, and AWS cloud computing. These businesses reinforce one another in some ways but have very different economics. Retail supplies enormous transaction volume, while AWS contributes a disproportionate share of operating profit.

In Q2, revenue grew **19.6%** and operating income **43.2%**. AWS produced approximately **61% of consolidated operating profit** on **21% of revenue**. The less comfortable result is cash: Amazon generated **$161.4 billion of trailing operating cash flow**, but its defined net capital spending exceeded that amount, leaving **negative $7.6 billion of FCF**. [Q2 release and financial statements][release]

Reported net income of **$62.6 billion** also included large non-operating investment gains. The central debate is therefore not whether Amazon has valuable businesses. It is whether growth in those businesses can produce attractive cash returns after infrastructure, logistics, and other investment, and how long investors must wait for that cash to appear.

## 1. How the business works

Amazon reports **North America**, **International**, and **AWS**. The retail geographic segments include several business models, including first-party product sales, third-party seller services, advertising, and subscriptions. Advertising is a separate revenue category, not a fourth operating segment. Adding advertising revenue to the geographic totals would count some revenue twice. [Annual business and segment notes][annual]

![Amazon Q2 2026 revenue composition; exact categories and values follow in the table.](https://stuff.ac/api/v1/items/f6ce4000-31df-4cfe-85b5-f25d05679efe/content)

| Revenue category | Revenue ($B) | Share |
| --- | ---: | ---: |
| North America | 116.177 | 57.9% |
| International | 42.197 | 21.0% |
| AWS | 42.232 | 21.1% |

*Q2 2026. Source: [quarterly segment statement][quarter]. The chart shows reported revenue segments, not separate retail product categories.*

When Amazon sells an item from its own inventory, it recognizes the product sale and carries the cost of the goods. When an independent seller uses the marketplace, Amazon generally earns fees for marketplace and related services. These revenue streams have different gross-versus-net presentation and different margins. A shift toward seller services can improve economics without a proportionate increase in reported revenue.

Fulfillment creates both differentiation and cost. Inventory must be positioned, goods moved, packages delivered, and returns handled. Higher local order density can improve efficiency, while faster delivery may increase demand. But speed has to earn its cost. Better customer experience alone does not prove that every incremental delivery option is profitable.

AWS sells computing, storage, databases, and other technology services. It can benefit as customers move workloads to the cloud and build AI applications. Its economics depend on utilization, pricing, equipment cost, power, operating efficiency, and the pace at which infrastructure must be replaced or expanded. The comparison with retail is useful precisely because the profit and cash profiles differ so much.

## 2. Revenue and profit growth

![Amazon annual revenue rises from $574.8B to $716.9B. Operating income and GAAP net income are shown separately; exact numbers follow.](https://stuff.ac/api/v1/items/b5395bcd-4c49-480c-8816-5fdebda6346e/content)

| Fiscal year / end | Revenue | Operating income | Net income | Operating margin |
| --- | ---: | ---: | ---: | ---: |
| FY2023 / 2023-12-31 | 574.785 | 36.852 | 30.425 | 6.4% |
| FY2024 / 2024-12-31 | 637.959 | 68.593 | 59.248 | 10.8% |
| FY2025 / 2025-12-31 | 716.924 | 79.975 | 77.670 | 11.2% |

| Year-over-year change | Revenue growth | Operating-income growth | Net-income growth |
| --- | ---: | ---: | ---: |
| FY2024 | 11.0% | 86.1% | 94.7% |
| FY2025 | 12.4% | 16.6% | 31.1% |

*Sources: [FY2025 annual statements][annual], [SEC financial facts][facts]. Years end December 31. Growth and margins are author calculations.*

Between 2023 and 2025, operating margin expanded from 6.4% to 11.2%. That is a substantial improvement across a large revenue base. Because much of Amazon's revenue comes from relatively low-margin retail activity, even a small change in consolidated margin can produce a large change in operating profit.

| Quarter / end | Revenue | Operating income | Net income | Operating margin | FCF |
| --- | ---: | ---: | ---: | ---: | ---: |
| Q3 2025 / 2025-09-30 | 180.169 | 17.422 | 21.187 | 9.7% | 1.297 |
| Q4 2025 / 2025-12-31 | 213.386 | 24.977 | 21.192 | 11.7% | 15.990 |
| Q1 2026 / 2026-03-31 | 181.519 | 23.852 | 30.255 | 13.1% | -17.202 |
| Q2 2026 / 2026-06-30 | 200.606 | 27.461 | 62.647 | 13.7% | -7.689 |

*Sources: [SEC facts][facts], [Q3 2025 release][q325], [annual statements][annual], and [Q2 2026 statements][quarter]. Q4 uses annual minus nine-month figures where appropriate.*

The quarterly history contains material discrete items. Q3 2025 included a **$2.5 billion FTC-related settlement charge** and **$1.8 billion of estimated severance costs**, which depressed operating income. In Q2 2026, management identified approximately **$1.2 billion of operating benefits**, split between tariff refunds in North America and energy-derivative gains in AWS. Comparing the quarters without these facts can exaggerate the improvement in a normal operating run rate. [Q3 release][q325], [Q2 call][call]

The latest quarter also included **$53.4 billion of non-operating income, net**, largely associated with investment gains. That explains much of the extraordinary net-income growth. Operating profit is the more useful first measure of customer-supported business performance; cash flow then shows how much of it remains after the buildout. [Q2 income statement and investment notes][quarter]

## 3. Where growth and profit come from

Q2 North America revenue was **$116.2 billion**, International **$42.2 billion**, and AWS **$42.2 billion**. Operating income was $9.1 billion, $1.7 billion, and $16.6 billion, respectively. Calculated segment operating margins were approximately **7.9%**, **4.1%**, and **39.4%**. AWS's importance to group profit is therefore much larger than its share of revenue suggests. [Q2 segment financials][release]

The AWS margin includes the energy-derivative benefit noted above. Even so, the segment's profit contribution is substantial. Its revenue growth reflects demand for both conventional cloud services and AI infrastructure. Treating all AWS expansion as a new AI revenue stream would overstate what Amazon separately discloses.

Retail improvement can come from product mix, third-party services, advertising, logistics efficiency, and better use of existing capacity. Those effects interact. More seller selection can attract shoppers; more shoppers can attract advertisers; higher density can reduce delivery cost per order. But if fees become unattractive to sellers or advertising makes the shopping experience worse, the same relationships can weaken.

Advertising monetizes access to shoppers near a purchasing decision. That can be valuable, but advertising performance should be judged by returns to advertisers, not only by the platform's ability to sell more placements. Amazon does not disclose a standalone advertising operating-profit segment, so this report does not invent an advertising margin or assign all retail improvement to ads.

For analysts, reported revenue is only the beginning of the comparison. A dollar of first-party product sales is economically different from a dollar of marketplace fees or cloud services. Consolidated revenue multiples are particularly blunt for a company with this mixture of activities.

## 4. Cash, capital spending, and financial quality

![Amazon cash from operations versus free cash flow across three completed fiscal years and latest TTM. Latest trailing FCF is $-7.6B; exact numbers follow.](https://stuff.ac/api/v1/items/d772b31f-cb45-4f98-91a5-04e6573912a8/content)

| Period / end | Operating cash flow | Cash capex | Free cash flow |
| --- | ---: | ---: | ---: |
| FY2023 / 2023-12-31 | 84.946 | 48.133 | 36.813 |
| FY2024 / 2024-12-31 | 115.877 | 77.658 | 38.219 |
| FY2025 / 2025-12-31 | 139.514 | 128.320 | 11.194 |
| TTM / 2026-06-30 | 161.403 | 169.007 | -7.604 |

*Amazon FCF = operating cash flow minus purchases of property and equipment, plus proceeds from property/equipment sales and incentives. Capex in this table is that net cash amount. Sources: [annual statement][annual], [quarterly cash-flow statement][quarter], [Q3 2025 release][q325]. TTM ends June 30, 2026.*

Trailing operating cash flow was strong, but **$169.0 billion of net capital spending** more than absorbed it. This is not the same as an operating cash loss. Amazon's existing activities generated cash, while the company invested even more. The investment can be valuable, but its size makes utilization, timing, and financing central to the investment case.

In Q2 alone, cash purchases of property and equipment were **$54.2 billion**. After $1.1 billion of related proceeds and incentives, the net figure was $53.1 billion against $45.4 billion of operating cash flow. Quarterly FCF was therefore **negative $7.7 billion**. The distinction between gross and net capital spending is explicit so comparisons do not silently mix definitions. [Q2 cash-flow statement][quarter]

Amazon's FCF measure does not deduct finance-lease principal or principal repayments on financing obligations. Those payments remain economic cash requirements. Nor does this FCF measure include acquisitions or investment purchases. A reader assessing all cash available to shareholders should inspect those additional uses rather than treating the headline figure as unrestricted distributable cash.

Working capital is another source of movement. A retailer collects customer cash and pays suppliers on different schedules. Growth can release cash when supplier payments follow customer receipts, while inventory accumulation or slower sales can reverse that benefit. Holiday seasonality makes individual quarters especially poor candidates for simple annualization.

First-half stock-based compensation expense was **$10.1 billion**, while the June balance sheet reported **$123.0 billion of cash, equivalents and marketable securities** and **$128.9 billion of long-term debt**, excluding separate lease liabilities and other claims. Long-term debt had increased substantially from year-end. These figures show why funding deserves attention even for a company with enormous operating cash generation. [Quarterly balance sheet and compensation note][quarter]

Investment gains require a separate analysis. A rising valuation for a model developer can increase Amazon's reported profit and asset values without paying for Amazon's current construction program. Selling an investment might release cash later, but that decision could have tax, strategic, and liquidity consequences. There is no assumption here that paper gains are immediately available to fund capex.

## 5. Product outlook and developments since earnings

AWS's approach includes a choice of third-party models, its own technology, and infrastructure based on both custom chips and outside suppliers. Amazon's product announcement says **Claude Fable 5.1 became available through Bedrock in September**. This establishes product availability through the platform, not AWS's margin on the service or proof of superiority over alternatives. [Bedrock product announcement, updated September 1][product]

Model choice can be commercially useful because customers may prefer different models for cost, quality, latency, or particular tasks. AWS can earn revenue even when it does not supply the winning model itself. However, that revenue has to cover the model provider's economics and the underlying computing cost. More choice does not guarantee an attractive retained margin.

In **August**, AWS and NVIDIA announced plans for **two million additional GPUs** and related infrastructure for 2027–2028. These are future deployment plans, not units already delivered or recognized AWS revenue. The announcement also demonstrates that custom silicon and NVIDIA infrastructure can coexist within the same platform. [AWS/NVIDIA announcement][gpu]

On **October 2**, Amazon announced more than **$1 billion over five years** for communities hosting data centers through its Built Together program. Associated Press reporting placed the announcement in the context of local resistance to data-center expansion. [Amazon announcement][community], [AP, October 2][news] The business implication is that permitting, power, water, and community acceptance influence when contracted capacity can become productive.

Retail AI tools can improve discovery, seller workflows, customer service, and internal operations. The commercial test is whether they improve conversion, retention, or cost per transaction after their own costs. Product demonstrations and internal usage anecdotes are useful leads for research, but this report does not assign undisclosed revenue or margins to them.

## 6. Competitive position

Amazon competes with retailers, marketplaces, logistics providers, advertising platforms, and cloud providers. The relevant advantage depends on the business. In retail, selection, price, convenience, delivery reliability, and customer trust matter. In cloud, breadth of services, developer familiarity, enterprise relationships, performance, and cost matter.

Scale can improve economics, but it can also make mistakes expensive. Building ahead of demand may improve availability when customers need capacity. It may also leave underused facilities or equipment if forecasts prove optimistic. The question is not whether management should ever invest ahead of demand; it is how much confidence the commitments justify and how much flexibility remains.

Amazon's relationship with AI partners creates both opportunity and concentration. A partner can be an investment, a customer, and a supplier of models offered through AWS. Those roles should be analyzed separately to avoid counting the same success multiple times. Investment appreciation, booked cloud commitments, recognized revenue, and collected cash are different stages of value creation.

Custom processors can offer cost or supply advantages for suitable workloads. Outside processors may remain attractive for compatibility, performance, or customer preference. It is unnecessary to assume a single winner across all computing tasks. The more useful test is whether AWS can offer a compelling service while retaining enough margin to pay for the infrastructure behind it.

## 7. What the earnings call clarified

The **July 30 prepared remarks and Q&A** put management's 2026 capex expectation at approximately **$220 billion**, citing higher component costs, and described AWS backlog of **$496 billion**. Those figures represent a spending expectation and future contracted business, respectively. Neither is current-year recognized revenue. [Q2 call][call]

Analyst questions explored funding and supply-chain inflation. Management discussed long-term agreements and the distinction between contracted and on-demand activity. The unanswered issue is how much cost inflation Amazon can recover through pricing as capacity arrives and agreements renew. Backlog improves visibility but cannot, on its own, settle that margin question. [Q&A on capital and contracts][call]

The next-quarter revenue guidance was **$197–202 billion**, with operating income of **$22.5–26.5 billion**, as of July 30. Prime Day occurred in June 2026 versus July 2025, shifting the comparison between Q2 and Q3. These are dated management forecasts and calendar effects, not a reason to assume the underlying demand trend changed by the same amount as reported quarterly growth. [Q2 release, guidance][release]

## 8. Valuation and expectations

The preserved October 6 premarket snapshot values Amazon at **$2.711 trillion**. This is company-wide equity value, not enterprise value. [Market-cap provider][market]

| Measure | Trailing denominator ($B) | Implied ratio |
| --- | ---: | ---: |
| Revenue | 775.680 | 3.5× |
| GAAP net income | 135.281 | 20.0× |
| Free cash flow | -7.604 | Not meaningful: negative FCF |

*Trailing period ends 2026-06-30. Ratios = dated equity value / stated trailing denominator. FCF yield = FCF / equity value = -0.28%. These approximate aggregate multiples use the provider's rounded market capitalization, not a quote multiplied by a newly estimated share count. Sources: [market snapshot][market], [reported financials][annual], [SEC facts][facts].*

An approximate **20.0× trailing GAAP earnings** multiple can make Amazon look less demanding than companies with higher P/E ratios. That comparison is incomplete because investment gains inflate the denominator and FCF is negative. A price/FCF multiple is therefore marked **not meaningful**, rather than presenting a negative number as evidence of cheapness.

Equity value divided by revenue is approximately **3.5×**, but Amazon's retail-heavy revenue base has different margins from a software business. A lower sales multiple does not automatically imply a cheaper business. A valuation should estimate the cash generated by each activity after the investment needed to sustain it, while reconciling corporate costs, debt, leases, and investment assets.

The attractive case is that today's buildout creates infrastructure capable of supporting profitable growth for years. The difficult case is that equipment must be replaced or expanded so rapidly that cash spending remains persistently high. Both are consistent with strong near-term revenue. The financial statements need to demonstrate which economics actually emerge.

## 9. Operating scenarios and risks

| Illustrative path | Revenue growth | Operating margin | Revenue ($B) | Operating income ($B) |
| --- | ---: | ---: | ---: | ---: |
| Lower | 8% | 10% | 837.7 | 83.8 |
| Middle | 13% | 12% | 876.5 | 105.2 |
| Higher | 18% | 14% | 915.3 | 128.1 |

The lower case assumes slower cloud and retail growth with pressure on margins. The middle case assumes growth continues while consolidated profitability stays near the trailing level. The higher case requires stronger AWS and retail contribution with good utilization. These are author-created next-twelve-month operating sensitivities, not management guidance or probability-weighted forecasts.

Capital spending is deliberately left outside the operating-profit scenarios because it requires its own assumptions. For example, $180 billion of operating cash flow and $200 billion of net cash capex would produce negative $20 billion of FCF under the defined formula. That is illustrative arithmetic, not a forecast. Raising the revenue-growth assumption alone does not resolve the cash question.

Risks include overbuilding capacity, delayed power or permits, component inflation, competitive pricing, retail execution, working-capital reversals, and partner concentration. Investment losses can also reduce net income independently of operating performance. Conversely, a moderation in capex could improve near-term FCF while signaling weaker future growth. Neither direction should be judged without context.

## 10. What to watch next

- **AWS economics:** revenue, margin after discrete benefits, customer collections, and utilization of new capacity.
- **Retail profitability:** North America and International operating income alongside shipping and fulfillment demands.
- **Cash:** trailing operating cash flow, net cash capex, lease payments, and financing needs.
- **AI delivery:** conversion of announced capacity and backlog into recognized, collected revenue.
- **Investment effects:** operating results separated from model-company valuation gains and losses.

The business case strengthens if AWS and retail improvement eventually produce more cash after investment. It weakens if reported earnings remain buoyant while capital requirements and financing obligations rise faster than the cash customers generate.

## Glossary and calculation notes

**First-party sales** mean Amazon sells goods itself. **Third-party services** mean Amazon earns fees from independent sellers. **AWS** is Amazon Web Services. **Backlog** is contracted future business under the company's definition. **Net cash capex** here subtracts proceeds and incentives from PPE purchases. **FCF** uses the company's stated cash formula and is not total cash after every investment or financing payment.

Growth = current period / prior comparable period − 1. Operating margin = operating income / revenue. TTM is the sum of the latest four quarters through 2026-06-30; standalone fourth-quarter and cash-flow values are derived from annual/YTD differences where necessary. Scenario revenue = TTM revenue × (1 + assumed growth); scenario operating profit = scenario revenue × assumed margin. Figures use reported US GAAP unless explicitly labeled otherwise. Percentages use unrounded data; displayed values can differ slightly through rounding.

Product announcements show stated availability and company claims. Forecasts retain their announcement dates. No consensus estimates, undisclosed product margins, or scenario probabilities are invented. This is a dated business analysis, not a live market feed.

## Further reading and sources

1. [Amazon FY2025 Form 10-K][annual] — annual statements, business models, segment economics, leases, commitments, and risks.
2. [Amazon Q2 2026 Form 10-Q][quarter] — quarter ended June 30. Investment gains, segment profit, cash spending, debt, and compensation.
3. [Q2 2026 earnings release][release] — July 30. Reported performance, cash reconciliation, and dated Q3 guidance.
4. [July 30 earnings-call transcript, reproduced by Stock Analysis][call] — prepared remarks and Q&A on capex, funding, contracts, and input-cost inflation.
5. [Q3 2025 earnings release][q325] — October 30, 2025. Historical standalone cash flows and the settlement/severance effects needed to interpret the four-quarter record.
6. [Claude on Amazon Bedrock][product] — updated September 1, 2026. Product availability; company claims, not independent model benchmarking.
7. [AWS/NVIDIA infrastructure announcement][gpu] — August 2026. Future 2027–2028 deployment plans, not completed installations.
8. [Amazon's Built Together announcement][community] — October 2, 2026. Planned community spending and the company's account of infrastructure constraints.
9. [AP on the community program][news] — October 2, 2026. Independent context on resistance to data-center development.
10. [SEC company financial facts][facts] — financial cross-checks restricted to filings available at the research cutoff.
11. [CompaniesMarketCap technology ranking][market] — October 6 premarket snapshot retained from the space README.

[annual]: https://www.sec.gov/Archives/edgar/data/1018724/000101872426000004/amzn-20251231.htm
[quarter]: https://www.sec.gov/Archives/edgar/data/1018724/000101872426000026/amzn-20260630.htm
[release]: https://ir.aboutamazon.com/news-release/news-release-details/2026/Amazon-com-Announces-Second-Quarter-Results/default.aspx
[call]: https://stockanalysis.com/stocks/amzn/transcripts/657334-q2-2026/
[q325]: https://ir.aboutamazon.com/news-release/news-release-details/2025/Amazon-com-Announces-Third-Quarter-Results/default.aspx
[product]: https://www.aboutamazon.com/news/aws/claude-fable-5-anthropic-available-amazon-bedrock
[gpu]: https://press.aboutamazon.com/aws/2026/8/aws-and-nvidia-to-deliver-2-million-additional-gpus-and-next-generation-infrastructure-for-agentic-and-physical-ai
[community]: https://www.aboutamazon.com/news/company-news/amazon-data-centers-built-together
[news]: https://apnews.com/article/91b65ba1729540c1c0d92e35deef35d8
[facts]: https://data.sec.gov/api/xbrl/companyfacts/CIK0001018724.json
[market]: https://companiesmarketcap.com/tech/largest-tech-companies-by-market-cap/