An investor’s argument with himself about AWS, retail margins, AI spending and whether Amazon’s profit machine is becoming stronger—or merely more expensive
Amazon has spent most of its public life teaching investors not to judge it like a normal company.
Whenever the profits looked thin, the explanation was that Amazon was investing. Whenever spending looked reckless, the explanation was that Amazon was building infrastructure everyone else would eventually need. Whenever Wall Street asked when the harvest might begin, Amazon planted another forest.
Annoyingly, this strategy worked.
The company turned an online bookstore into a retail empire, a logistics network, an advertising platform, a subscription ecosystem and the world’s leading cloud-infrastructure business. It built warehouses when skeptics wanted margins, servers when analysts wanted discipline and delivery capacity when investors were still trying to understand why anyone needed a package in less than two days.
Now the argument has changed. Amazon is no longer being asked whether it can make money. The company produced $27.5 billion in operating income in the second quarter of 2026, up 43% from a year earlier. The question is whether this enormous machine can keep becoming more profitable while simultaneously pouring extraordinary sums into artificial intelligence, custom chips, data centers, robotics and satellites.
That is a much more interesting question.
As I look at Amazon today, I can construct a convincing bull case before breakfast and an equally persuasive bear case by lunch. The numbers are strong enough to inspire confidence and complicated enough to punish anyone who reads only the headline. Amazon’s operations are producing more profit, but its capital requirements are devouring cash. AWS is accelerating, but the AI arms race is becoming financially medieval. Retail margins are improving, but consumers remain price-sensitive and competition has not taken a sabbatical.
So can Amazon keep expanding profitability?
My short answer is yes—but not smoothly, cheaply or indefinitely. My longer answer requires allowing the bull and the bear to finish their drinks.
The Latest Numbers Are Not Subtle
Amazon’s second-quarter 2026 results gave the optimists plenty to work with. Net sales rose 20% year over year. AWS sales increased 37% to $42.2 billion. Companywide operating income climbed from $19.2 billion to $27.5 billion.
The segment breakdown was just as important. North American operating income rose to $9.1 billion from $7.5 billion. International operating income increased to $1.7 billion from $1.5 billion. AWS operating income surged to $16.6 billion from $10.2 billion.
Those figures describe a company whose profit engine is not confined to a single lucky corner. The cloud business is thriving, the domestic retail operation is more efficient and the international segment is no longer behaving like a charitable foundation for global package delivery.
Then there is net income: $62.6 billion, or $5.75 per diluted share.
That number is magnificent and largely useless as a measure of recurring quarterly profitability because it included $53.4 billion of pre-tax non-operating income, primarily related to Amazon’s investment in Anthropic. This is why I do not fall in love with earnings headlines. Accounting occasionally arrives wearing a tuxedo and carrying a fog machine.
The Anthropic gain is economically meaningful. Amazon owns an asset that appreciated substantially. I would rather the investment rise than collapse. But it did not come from selling more cloud capacity, delivering more packages or improving advertising margins. For evaluating the operating business, I focus on operating income and cash generation.
And this is where the story becomes less comfortable.
Trailing-12-month free cash flow fell to an outflow of $7.6 billion, compared with an inflow of $18.2 billion a year earlier. Amazon attributed the decline primarily to a $66.1 billion year-over-year increase in purchases of property and equipment, largely reflecting AI investments. Management said earlier in 2026 that it expected roughly $200 billion in capital expenditures for the year.
Two hundred billion dollars is not a capital budget. It is a national weather system.
The central debate is therefore easy to state: Is Amazon buying the infrastructure that will produce the next decade of high-margin growth, or is it entering a permanent cycle in which every dollar of AI opportunity requires another dollar to be fed into an increasingly expensive furnace?
The Bull Case: AWS Is Reaccelerating When It Matters Most
The strongest argument for expanding profitability begins with AWS.
For years, AWS has been Amazon’s economic masterpiece. Retail made Amazon indispensable to consumers, but cloud computing gave it the margins investors had spent decades waiting for. In the second quarter of 2026, AWS generated about 60% of Amazon’s total segment operating income while accounting for far less than 60% of company sales.
That is what a premium business looks like.
More importantly, AWS growth accelerated to 37%. In a company Amazon’s size, that is not merely healthy growth. It is a small economy discovering espresso.
The AI boom is creating enormous demand for computing capacity, model training, inference, data storage, databases, networking and specialized chips. Amazon does not need to predict which single AI application will dominate. It can profit by renting the digital picks and shovels to thousands of companies trying to find out.
This is the classic platform advantage. The winners, losers, visionaries and spectacularly funded nonsense can all consume AWS resources.
Amazon also has several strategic levers. It offers Nvidia-based infrastructure while developing its own Trainium and Inferentia chips. Custom silicon can potentially lower costs, improve performance and reduce dependence on one supplier. Bedrock gives enterprise customers access to multiple foundation models rather than forcing every company to pledge loyalty to one AI religion. Amazon’s relationship with Anthropic strengthens the ecosystem while creating demand for AWS infrastructure.
If AI becomes a foundational computing layer rather than a temporary speculative fever, AWS may be entering a new growth cycle. Revenue growth can support operating leverage because data centers contain substantial fixed costs. Once capacity is built and utilization rises, incremental revenue can carry attractive margins.
The bull sees today’s capital spending as tomorrow’s toll road.
I understand that argument because Amazon has done this before. The company’s history is a long sequence of investments that appeared excessive until competitors wished they had made them first. Fulfillment centers looked like margin destruction until they became a logistics moat. AWS looked like an odd internal project until it became the profit center supporting the empire. Prime looked expensive until it changed consumer behavior.
Betting against Amazon’s willingness to invest has not been a relaxing hobby.
The Bull Case: Retail Is Finally Behaving Like a Business
The second pillar of the bull case is less glamorous but equally important: Amazon’s retail operations are becoming more efficient.
For a long time, Amazon retail resembled a breathtaking feat of engineering operated for the emotional benefit of customers and the financial confusion of shareholders. The company moved staggering volumes of goods while producing margins so thin they could hide behind dental floss.
That has changed.
Amazon redesigned its U.S. fulfillment network around regionalization, placing inventory closer to customers and reducing the distance packages travel. Faster delivery is excellent for consumers, but the investor benefit is cost efficiency. Shorter routes can reduce transportation expense, improve asset utilization and increase the number of orders Amazon can handle without proportionally increasing costs.
The North America segment produced $9.1 billion in operating income during the second quarter. International generated $1.7 billion. Those profits suggest the company is extracting more value from infrastructure already built.
Retail margin expansion does not require Amazon to become a luxury-goods company. Small improvements applied to hundreds of billions in sales can produce enormous incremental profit. A percentage point that looks trivial on a presentation slide can become billions of dollars when applied to Amazon’s revenue base.
Several forces could continue helping:
Better inventory placement and delivery density can lower fulfillment costs.
Automation and robotics can reduce repetitive labor and improve throughput.
More third-party sellers can increase selection while allowing Amazon to collect fees without owning every unit of inventory.
Prime membership deepens customer loyalty and supports recurring revenue.
Faster delivery can increase purchase frequency, spreading fixed infrastructure costs across more orders.
The bull does not need retail margins to become spectacular. The bull needs them to become consistently less mediocre.
Amazon has spent decades building a network competitors cannot easily reproduce. If management can make that network incrementally more productive every year, the profit contribution from retail can remain substantial even at modest margins.
The Bull Case: Advertising Is the Quiet Margin Machine
Amazon advertising may be the most underappreciated part of the profitability story.
When consumers search Amazon, they are frequently close to making a purchase. That gives the company something advertisers value enormously: intent. A social-media user may be looking at vacation photos or arguing with a cousin. An Amazon user searching for a cordless drill has announced a commercial purpose with admirable clarity.
Sponsored product placements and other advertising formats monetize that intent. The underlying economics can be attractive because Amazon already owns the marketplace, customer traffic and transaction data. It does not need to build an entirely separate audience before selling access to it.
Advertising also strengthens the broader ecosystem. Sellers pay for visibility, Amazon earns high-margin revenue, consumers receive product suggestions of varying usefulness and the marketplace becomes more monetized without Amazon having to purchase additional inventory.
There is a limit, of course. If every search result becomes an advertisement, the customer experience begins resembling a bazaar operated by slot machines. Amazon must balance monetization with relevance. Still, the runway appears meaningful across sponsored listings, streaming video, connected television and other media properties.
The bull case is not simply “AWS grows.” It is that Amazon possesses several businesses with very different economics. Retail supplies scale and customer relationships. Prime increases loyalty. Advertising monetizes purchase intent. AWS supplies high-margin infrastructure. Logistics creates a moat. Together, these operations can reinforce one another.
Few companies have that many levers available at once.
The Bull Case: Operating Leverage Has Already Arrived
I am naturally suspicious of any investment thesis built entirely on future margin expansion. “Profits will arrive later” has financed some of capitalism’s most beautifully designed bonfires.
Amazon’s case is stronger because the improvement is already visible.
Operating income rose from $36.9 billion in 2023 to $68.6 billion in 2024 and $80.0 billion in 2025. The second quarter of 2026 added another clear step upward. This is not a hypothetical PowerPoint margin. It is reported operating profit.
Management has demonstrated greater cost discipline since the pandemic-era overexpansion. Head-count reductions, network optimization, slower growth in certain expenses and improved international performance have changed the earnings profile. The company is proving that revenue can grow faster than some operating costs.
That is the essence of operating leverage.
The optimistic interpretation is that Amazon has entered a new phase. It still invests aggressively, but the existing businesses are mature enough to throw off increasing operating income while new bets are funded. Under this view, occasional periods of heavy capital spending do not invalidate margin progress; they temporarily mask the cash returns of an increasingly profitable system.
If AWS growth remains elevated, advertising expands and retail efficiency continues improving, Amazon could generate rising operating margins even if the year-to-year path includes turbulence.
That is a formidable bull case.
Now I have to ruin the mood.
The Bear Case: AI Has Turned Capital Discipline Into a Historical Document
The bear begins with the $200 billion capital-spending plan and asks a question bulls dislike: What return will Amazon actually earn on all that money?
AI infrastructure is expensive, depreciates and can become technologically obsolete faster than management presentations imply. Data centers require chips, power, cooling, networking equipment, land and construction. Then newer chips arrive promising better performance, customers demand lower prices and competitors build their own capacity with comparable enthusiasm.
Amazon is not investing in isolation. Microsoft, Alphabet, Meta and others are spending heavily. Each company believes demand is enormous. Each company believes its infrastructure is differentiated. Each company possesses executives capable of saying “unprecedented opportunity” without blinking.
They cannot all earn extraordinary returns on unlimited capital forever.
The danger is not that AI demand disappears. The danger is that demand grows while economics disappoint. Industries can experience explosive usage and miserable returns simultaneously. Airlines transport enormous numbers of passengers. That has not made every airline investor serene.
Cloud providers may face price competition, rising power costs, hardware depreciation and customers becoming more efficient with inference. Some large customers will build more internal infrastructure. AI laboratories may negotiate aggressively because their workloads are massive. If supply eventually catches demand, utilization and pricing could weaken.
Meanwhile, the free-cash-flow outflow is real. Amazon’s operations are producing substantial cash, but capital spending is consuming even more. Bulls call this investment. Bears call it an invoice.
Both are correct until the returns arrive.
The Bear Case: AWS Carries an Uncomfortable Amount of the Profit
AWS generated $16.6 billion of segment operating income in the second quarter, compared with Amazon’s total segment operating income of $27.5 billion. That concentration is a strength when AWS is accelerating. It becomes a vulnerability if cloud growth or margins stumble.
Retail profits are improving, but they remain structurally lower-margin and exposed to labor, fuel, shipping, wage and inventory costs. Advertising is attractive but intertwined with marketplace activity and regulatory scrutiny. AWS is still the portion of the company doing the heaviest financial lifting.
The bear asks what happens if AWS growth normalizes after the current AI buildout, or if capital intensity remains permanently higher. Revenue growth alone does not guarantee expanding profitability. AWS operating income must grow faster than the costs required to support it.
There is also the problem of customer concentration and AI economics. The largest model developers consume huge amounts of computing power, but some also possess the ambition and financing to build their own systems. Amazon’s Anthropic partnership creates strategic alignment, yet it also tangles investment gains, cloud commitments and competitive positioning in ways that require more careful analysis than “AI equals good.”
The cloud market remains attractive, but Microsoft Azure and Google Cloud are not operated by distracted amateurs. Competition will remain intense across models, chips, developer tools and enterprise relationships.
Amazon can win a large share of a growing market and still experience margin pressure. Investors often forget that two things can happen at once because spreadsheets prefer cleaner stories.
The Bear Case: Retail Efficiency Has Natural Limits
The retail business has become more profitable, but I would not project recent margin gains into eternity.
There are only so many fulfillment routes to regionalize. Only so many excess costs to remove. Eventually efficiency initiatives move from dramatic redesign to incremental improvement. The easiest savings are harvested first; later gains require greater effort.
Amazon also competes on price and convenience. Those customer promises are not free. Faster delivery requires inventory distributed across more locations, local capacity and sophisticated forecasting. Quick commerce increases complexity. International expansion can require lower prices and substantial investment. Amazon Leo, the company’s satellite-network initiative, adds another expensive project whose financial returns remain uncertain.
Labor costs will continue rising over time. Regulatory pressure may affect the marketplace, worker classification, seller practices, advertising or Prime. Antitrust authorities in multiple jurisdictions are not examining Amazon because they admire the website layout.
Competition is also changing. Walmart has improved its digital capabilities and can combine stores with fulfillment. Chinese platforms pressure prices and reshape consumer expectations. Shopify supports merchants seeking alternatives. Retailers are learning that ignoring e-commerce is not a strategy, although some discovered this at a pace normally associated with continental drift.
Amazon’s scale remains a moat, but moats require maintenance. The retail network must be funded, staffed, upgraded and defended.
Margin expansion may continue, but it will not proceed as though gravity signed a waiver.
The Bear Case: The Stock Already Knows Amazon Is Excellent
A wonderful company can be a disappointing investment if the purchase price assumes too much wonder.
As of August 10, 2026, Amazon shares were around $274.48, giving the company a market capitalization near $3.0 trillion. The quoted price-to-earnings ratio was roughly 22, but that figure was distorted by the enormous non-operating Anthropic gain included in earnings. Using headline net income without adjustment makes the valuation look cheaper than the recurring business warrants.
This is why I would not anchor on the displayed P/E ratio. I would evaluate Amazon using normalized operating earnings, expected future free cash flow and a range of outcomes for AI capital intensity.
At nearly $3 trillion, Amazon is not an undiscovered village bakery. Expectations are substantial. The market already understands that AWS is valuable, advertising is growing and retail margins have improved. Future returns depend not merely on business success but on success exceeding the assumptions embedded in the stock price.
The valuation can work if operating income compounds at a strong rate and capital spending eventually converts into much larger cash flow. It becomes vulnerable if AI infrastructure produces lower returns, AWS growth slows or free cash flow remains depressed longer than expected.
Investors should not ask only, “Will Amazon earn more?” The likely answer is yes. The better question is, “Will Amazon earn enough more, soon enough, to justify the price I am paying today?”
That question has fewer fans because it cannot be answered with a rocket emoji.
What I Think Happens Next
I believe Amazon can continue expanding operating profitability over the next several years, but the improvement will be uneven and cash flow will tell a less flattering story than operating income during the peak AI investment cycle.
My base case rests on four assumptions.
First, AWS remains a major beneficiary of AI demand. I do not expect 37% growth indefinitely, but I believe cloud infrastructure, model services, data tools and inference workloads can support healthy expansion. AWS has the scale, ecosystem and enterprise relationships to remain one of the central platforms.
Second, retail efficiency continues improving, though at a slower pace. Regional fulfillment, automation and delivery density should support margins, but I would not assume endless gains. The business remains competitive and operationally demanding.
Third, advertising continues to grow faster than the retail marketplace and contributes disproportionately to profit. Amazon’s proximity to the transaction creates an advantage that should remain valuable unless over-monetization damages the customer experience.
Fourth, capital intensity eventually moderates relative to operating cash flow. This is the most important assumption. If annual capital spending remains near $200 billion and continues rising as fast as demand, shareholders may spend years admiring operating profits they cannot fully collect. If the infrastructure buildout produces strong utilization and capex growth slows, free cash flow could rebound dramatically.
I am therefore bullish on Amazon’s ability to expand operating profit, cautiously optimistic about long-term free cash flow and unwilling to treat either outcome as guaranteed.
The Metrics I Would Watch
Rather than arguing endlessly about whether Amazon is “good” or “bad,” I would monitor a short list of evidence:
AWS revenue growth and operating margin. Strong growth matters, but margin reveals whether the growth is becoming more or less valuable.
Capital expenditures relative to operating cash flow. I want to see whether cash generation begins catching the spending cycle.
Trailing-12-month free cash flow. This will remain noisy, but the direction matters.
North America operating margin. Continued improvement would show that fulfillment efficiencies are durable.
International operating income. Consistent profitability would turn a historical drag into a meaningful contributor.
Advertising growth and customer experience. Monetization is attractive until search results become a paid obstacle course.
Depreciation expense. Today’s infrastructure spending becomes tomorrow’s expense. The bill does not disappear because the server looked impressive at the ribbon-cutting.
Management’s commentary on AI capacity and utilization. Persistent shortages can justify investment; excess capacity would change the thesis.
These metrics allow the thesis to evolve with evidence. I do not want to become emotionally loyal to either the bull or bear case. Stocks do not reward ideological consistency. They reward being approximately right before everyone else becomes precisely certain.
My Verdict
The bull case wins for me, but not by knockout.
Amazon has already demonstrated that profitability can expand across AWS, North America and international operations. AWS is reaccelerating at a strategically important moment. Advertising offers attractive economics. Retail efficiency has moved from promise to reported profit. The company possesses scale, infrastructure and customer relationships that few competitors can reproduce.
But the bear is not being pessimistic merely for recreational purposes. Free cash flow has turned negative on a trailing basis because Amazon is spending at a scale that would make previous versions of Amazon look financially restrained. AI may justify that investment, but the returns must eventually appear in cash, not only in revenue growth and executive enthusiasm.
The Anthropic valuation gain also reminds me to separate recurring operations from headline earnings. A $62.6 billion quarterly profit looks spectacular because it is spectacular. It is not a sensible quarterly run rate.
So can Amazon keep expanding profitability?
Yes, I believe it can. AWS, advertising and retail efficiency provide multiple paths. But I expect operating margins and free cash flow to diverge during the infrastructure buildout. The company may become more profitable on the income statement while appearing less generous in the cash-flow statement.
That tension is not necessarily a warning sign. It is the price of the thesis.
Amazon has always asked investors to tolerate present spending in exchange for future dominance. The difference now is the number of zeros. At nearly $3 trillion in market value, the company no longer receives unlimited credit merely for thinking long term. It must prove that this generation of investment produces returns worthy of its cost.
I would not bet casually against Amazon. The company has repeatedly turned expensive infrastructure into competitive advantage. But I would not excuse every expenditure simply because previous bets succeeded. History is evidence of capability, not a lifetime exemption from arithmetic.
For me, Amazon remains a high-quality business with a favorable long-term profitability outlook and a demanding near-term capital cycle. The bull case depends on operating leverage continuing. The bear case depends on capital intensity swallowing too much of the benefit. The winner will be decided not by how much AI capacity Amazon builds, but by how profitably customers use it.
That is the part of the story no keynote can settle.
Only the cash flow will.
Market data and company results are current as of August 10, 2026. This article expresses personal analysis for informational purposes and is not individualized investment advice.
Comments
Post a Comment