Microsoft Earnings Market Rally: Why Wall Street Rewarded AI Spending—and What Came Next

Wall Street’s sharp rebound on July 30, 2026, was not a simple return of risk appetite. It was a verdict on a question that has dominated the technology sector for much of the year: can the largest artificial-intelligence spenders convert record infrastructure budgets into revenue, operating profit, and cash quickly enough to justify the cost?

Microsoft gave investors the clearest affirmative answer yet. Its fiscal fourth-quarter results showed revenue of $90 billion, Azure growth of 43%, Microsoft Cloud revenue of $59.3 billion, and quarterly operating cash flow of $55.4 billion. The company spent $41 billion on capital expenditures during the quarter and still produced $19.6 billion of free cash flow. Microsoft shares rose more than 15% in regular trading on July 30, adding roughly $450 billion in market value and helping lift the Nasdaq Composite 2.78%, the S&P 500 1.66%, and the Dow Jones Industrial Average 1.19%.

That performance temporarily changed the market’s framing of the AI investment cycle. The central issue was no longer whether spending was enormous; that was already established. The more useful distinction became whether each company could show a credible chain from data-center construction to capacity, from capacity to customer usage, and from usage to cash generation. Microsoft demonstrated that chain more convincingly than Meta Platforms had the same evening. Amazon strengthened the bullish case after the close with accelerating AWS growth, while Apple offered a different model: strong product revenue and cash generation without matching the hyperscalers’ infrastructure intensity, but with supply constraints that disappointed investors.

The result was a powerful but narrow rally. Technology shares surged, semiconductor stocks rebounded, and Microsoft recorded its largest one-day percentage gain since 2008. Yet most S&P 500 constituents declined, the equal-weight index weakened, and long-term Treasury yields remained near levels last seen before the global financial crisis. In other words, the day contained two simultaneous messages. Investors were willing to reward companies that could prove AI spending was productive, but the cost of capital and the burden of financing that spending remained unusually high.

Last updated: July 31, 2026, 3:55 a.m. EDT. Market figures distinguish regular-session closes from after-hours reactions where relevant.

Key Takeaways

• Main development: Microsoft’s fiscal fourth-quarter results triggered a concentrated technology rally on July 30, 2026, after the company paired rapid Azure growth with positive free cash flow despite $41 billion of quarterly capital expenditure.
• Market response: Microsoft rose more than 15% in regular trading, adding approximately $450 billion in market value. The Nasdaq gained 2.78%, the S&P 500 rose 1.66%, and the Dow advanced 1.19%.
• Why it mattered: Investors had punished several AI-linked companies for rising capital intensity and weaker cash flow. Microsoft showed that large AI investment can be rewarded when revenue growth, capacity utilization, margins, and cash generation move together.
• Important qualification: The rally was not broad. Technology carried the indexes while a majority of S&P 500 stocks fell, indicating that the market rewarded specific earnings evidence rather than embracing a general risk-on move.
• Bond-market warning: The 30-year Treasury yield remained around 5.2%, near its highest level since 2007, after the Federal Reserve held its policy rate at 3.5% to 3.75% and three officials dissented in favor of an increase.
• What happened next: Amazon reported 37% AWS growth and a 43% increase in operating income, while Apple posted 16% revenue growth but issued a softer outlook linked to supply constraints. Amazon shares rose in after-hours trading; Apple shares fell.

Why the Stock Market Rallied After Microsoft Earnings

The immediate answer is straightforward: Microsoft exceeded expectations in the parts of its business most closely associated with the AI investment thesis, then gave investors reasons to believe the momentum could continue. Azure and other cloud-services revenue grew 43% from the comparable quarter, Intelligent Cloud revenue reached $39.3 billion, and management forecast approximately 45% constant-currency Azure growth for the next quarter. Demand continued to exceed available capacity, which meant new computing infrastructure was being monetized rapidly rather than sitting idle.

That last point was essential. Capital expenditure is not automatically bullish or bearish. Spending creates value when it expands a constrained, profitable business whose customers are willing to pay. It destroys value when companies overbuild, misjudge demand, or fund projects that earn returns below their cost of capital. Microsoft’s earnings suggested that its new data-center capacity was being absorbed almost as soon as it became available. Chief Financial Officer Amy Hood said efficiency gains across the CPU and GPU fleet, together with operational improvements that brought new capacity online sooner, contributed to Azure’s better-than-expected growth.

The market had entered the report skeptical. Alphabet had disclosed negative quarterly free cash flow after capital expenditure exceeded operating cash flow, and Meta’s second-quarter free cash flow fell to $784 million as its infrastructure bill expanded. Semiconductor stocks had sold off, long-duration growth shares were struggling with higher bond yields, and the market was questioning whether AI capital spending had moved too far ahead of commercial demand.

Microsoft did not eliminate those concerns, but it showed that they cannot be applied uniformly. Its quarterly capital expenditure of $41 billion was enormous, yet cash flow from operations rose 30% to $55.4 billion. Free cash flow remained positive at $19.6 billion. The company returned $10.2 billion to shareholders through dividends and repurchases during the quarter and more than $43 billion during the fiscal year. Investors therefore saw a company financing growth from an expanding operating engine rather than depending entirely on debt issuance, asset sales, or a reduction in shareholder returns.

That is why the rally spread most strongly through the semiconductor complex. Microsoft’s results supported the view that demand for CPUs, GPUs, networking, storage, and data-center equipment remained commercially productive. The Philadelphia Semiconductor Index jumped about 8.2%, with especially large moves in memory and processing names. The market was not merely celebrating a software-company earnings beat; it was repricing the probability that AI infrastructure demand could remain elevated without immediately collapsing customer economics.

The rally also reflected positioning. Technology and semiconductor stocks had already experienced a sharp drawdown, so a strong report from one of the industry’s largest buyers and sellers of AI capacity forced investors to reassess bearish trades. When a crowded concern meets better-than-feared evidence, price moves can exceed the change in fundamental value. Microsoft’s roughly $450 billion one-day market-cap gain illustrates that dynamic. It represented a dramatic shift in perceived risk, not $450 billion of cash suddenly entering the business.

The Rally Was Powerful, but It Was Not Broad

Headline indexes can conceal the distribution of returns. Microsoft has an unusually large weight in the S&P 500, Nasdaq Composite, and Dow Jones Industrial Average. A 15%-plus move in one of the world’s most valuable companies mechanically lifts market benchmarks even when the average stock is weak.

That is what happened on July 30. Technology rose more than 5% as a sector, but most S&P 500 companies declined. The equal-weight S&P 500, which gives each member the same influence, finished lower. Only a minority of Dow components advanced. The market therefore did not issue a broad endorsement of economic conditions, consumer demand, or corporate profit trends. It issued a focused endorsement of companies that could show credible AI monetization.

This distinction matters for interpreting the day. A broad rally usually suggests improving confidence across industries. A concentrated rally can instead represent a rotation into a small group of perceived winners. The latter is more fragile because index performance becomes dependent on a handful of companies continuing to deliver exceptional growth.

Concentration also affects valuation. When investors decide that one company has solved a problem its competitors have not, they may grant that company a premium multiple. Microsoft’s earnings created precisely that contrast. Meta reported strong revenue growth, but its costs rose much faster, its operating margin fell, and its free cash flow nearly disappeared. Microsoft reported strong growth while preserving positive cash flow. Both companies were spending heavily on AI, but the financial evidence looked different.

A narrow rally is not necessarily unhealthy. Markets are supposed to discriminate between business models. The concern arises when index gains are mistaken for confirmation that risk has declined everywhere. Long-term yields remained elevated, inflation remained above the Federal Reserve’s target, and parts of the economy continued to face financing pressure. Those conditions did not disappear because Microsoft executed well.

What Microsoft Actually Reported

Microsoft’s fiscal 2026 fourth quarter covered the three months ended June 30, 2026. Revenue was $90 billion, up 18% year over year. Operating income increased 18%, and adjusted diluted earnings per share rose 23% to $4.74 when excluding the impact of the company’s OpenAI investment. The company said annual revenue exceeded $331 billion, Microsoft Cloud revenue surpassed $214 billion, and Azure annual revenue exceeded $100 billion.

The most important operating metric was Azure growth. Revenue from Azure and other cloud services increased 43%, compared with a prior-year quarter that already contained accelerating growth. Customer demand still exceeded capacity, and management said additional capacity brought online during the quarter was monetized quickly.

Commercial remaining performance obligations, a measure of contracted revenue not yet recognized, increased 84% to $678 billion. Because a large OpenAI commitment influenced the total, Microsoft also disclosed that remaining performance obligations rose 25% excluding OpenAI. Roughly 30% of the total was expected to be recognized as revenue during the following 12 months. That mix matters: a large backlog can provide revenue visibility, but its duration, customer concentration, and cancellation terms affect how investors should interpret it.

Microsoft Cloud revenue reached $59.3 billion, up 27%. Cloud gross margin was 65%, down year over year because Azure represented a larger share of sales and the company continued to invest in AI infrastructure and product usage. The decline in gross-margin percentage illustrates a recurring tension in cloud economics. Revenue can grow rapidly while the mix shifts toward more infrastructure-intensive services that carry lower incremental margins during the buildout phase.

Source: Microsoft’s fiscal 2026 fourth-quarter earnings call and prepared financial commentary. Dollar figures are U.S. dollars and relate to the quarter ended June 30, 2026 unless otherwise stated.

Azure Growth Was the Strongest Evidence of AI Monetization

AI infrastructure spending is easier to defend when the customer-facing business is capacity constrained. Microsoft said demand exceeded supply, a statement that carries more weight when accompanied by 43% Azure growth and a forecast for approximately 45% constant-currency growth in the next quarter.

The economics can be understood in stages. First, Microsoft commits capital to land, buildings, power, networking equipment, CPUs, GPUs, and associated software. Second, it turns those assets into cloud capacity. Third, customers reserve or consume that capacity through Azure, Microsoft 365, GitHub, data services, and AI applications. Fourth, revenue must grow faster than depreciation, power, staffing, financing, and maintenance costs if the investment is to create durable value.

The July results suggested Microsoft was progressing through all four stages. Capacity additions supported revenue growth. Efficiency improvements allowed the company to bring infrastructure online sooner. Customer demand remained larger than available supply. Operating cash flow expanded even after the company absorbed higher lease payments and infrastructure costs.

Microsoft also benefited from a broad platform rather than one AI product. Azure sells compute and storage. GitHub Copilot monetizes developer activity. Microsoft 365 Copilot adds premium pricing to an installed base of commercial users. Database products such as PostgreSQL and Cosmos DB support agent workloads. The company’s model catalog includes third-party systems as well as its own models, reducing dependence on a single provider.

That diversification matters because AI demand can shift between training, inference, enterprise agents, coding, search, and productivity applications. A platform able to sell infrastructure and software across those categories has more ways to recover its investment. Microsoft is not relying solely on consumer subscriptions or a future advertising model. It is charging for capacity, usage, seats, premium features, and enterprise commitments.

The company said paid Microsoft 365 Copilot seats exceeded 30 million and more than doubled sequentially. It also reported stronger-than-expected GitHub Copilot consumption after changing the product’s business model to align pricing more closely with usage and value. These details suggest Microsoft is testing more than one monetization structure, an important step because flat per-seat pricing may not capture the economics of workloads whose compute costs vary widely.

Microsoft’s Cash Flow Was More Important Than Its Earnings Beat

Earnings per share can be influenced by investment gains, depreciation assumptions, tax effects, and stock-based compensation. Free cash flow is not immune to accounting choices, but it provides a more direct view of whether the company’s operations can finance its capital program.

Microsoft generated $55.4 billion of operating cash flow and $19.6 billion of free cash flow in the quarter. Capital expenditures totaled $41 billion, including $5.6 billion of finance leases, while cash paid for property and equipment was $35.8 billion. Roughly two-thirds of capital spending went toward short-lived assets, primarily CPUs and GPUs, rather than long-lived buildings and sites.

The short useful life of computing equipment is economically important. A data-center shell can operate for decades, but advanced accelerators can become commercially inferior much sooner. Rapid hardware improvement creates reinvestment risk: a company may need to replace or upgrade expensive equipment before the original asset has generated the expected return. High utilization and strong pricing are therefore essential during the early years of an AI server’s life.

Microsoft’s positive free cash flow indicated that it could fund a large portion of the buildout internally. That does not guarantee attractive returns. Free cash flow of $19.6 billion was lower than it would have been without the spending surge, and management expected fiscal 2027 capital expenditure to increase again. Still, the company entered the new fiscal year with an operating engine strong enough to absorb the investment while remaining cash-flow positive.

Management also changed the estimated useful life of data centers and office buildings from 15 years to 25 years. Microsoft said the accounting change would have only a minimal benefit to fiscal 2027 operating income, but it would affect the classification of future leases. More leases are expected to be treated as operating leases rather than finance leases, which means they will not appear in capital expenditures in the same way. The company adjusted its calendar 2026 capital-expenditure expectation to approximately $175 billion because of that reclassification.

Investors should not confuse a classification change with a reduction in economic commitment. An operating lease still creates payment obligations. The change can make reported capital expenditure less comparable across periods unless analysts also examine lease liabilities and cash payments. Microsoft was unusually explicit about this distinction, and it should remain part of any assessment of its AI spending trajectory.

Why Microsoft Was Rewarded While Meta Was Punished

Microsoft and Meta both reported rapid revenue growth and aggressive AI investment, but the market focused on the quality and timing of the return. Meta’s second-quarter revenue increased 28% to $60.8 billion, yet costs and expenses rose 55% to $42.0 billion. Operating income fell 8% to $18.8 billion, net income declined 14% to $15.8 billion, and operating margin dropped from 43% to 31%.

Meta spent $31.08 billion on capital expenditures, including finance-lease principal payments, and generated $31.86 billion of operating cash flow. The difference left only $784 million of free cash flow, down from $8.55 billion a year earlier. That result does not mean Meta’s AI strategy is failing. Revenue grew strongly, ad impressions increased 14%, and average price per ad rose 12%. It means the company’s current cash return on incremental infrastructure is less visible than Microsoft’s.

Meta’s business model also makes the comparison harder. It does not have a cloud platform of Microsoft’s scale where external customers directly purchase compute capacity. AI can improve ad targeting, engagement, content recommendations, creator tools, and new products, but the revenue link is indirect. The company must show that higher engagement or better ad performance produces enough incremental profit to cover the infrastructure bill.

Microsoft can sell the shovel and use the shovel. It operates its own AI applications while charging other companies to build on Azure. Meta primarily uses infrastructure to improve its own consumer and advertising ecosystem, although it has discussed broader enterprise opportunities. That structural difference helps explain why identical spending amounts would not deserve identical valuation treatment.

The market’s reaction therefore reflected evidence, not a blanket preference for one management team. Microsoft showed higher cloud growth, rising operating cash flow, positive free cash flow, and a large contracted backlog. Meta showed strong top-line growth but declining profit, sharply lower free cash flow, and a much larger increase in costs. The burden of proof shifted accordingly.

The AI Capex Era Is Replacing an Older Big Tech Playbook

For much of the past two decades, the largest technology companies were attractive partly because they combined high margins with modest physical capital needs. Software could be distributed to another customer at a low incremental cost. Advertising platforms converted user attention into revenue without building factories. Mature companies used excess cash to repurchase shares, raise dividends, acquire smaller competitors, or accumulate large balance-sheet reserves.

Generative AI has changed that model. Competitive advantage now depends on access to power, land, data-center capacity, networking equipment, advanced processors, memory, cooling systems, and specialized engineering talent. These are capital-intensive inputs. The largest platforms are becoming infrastructure operators on a scale that resembles telecommunications, utilities, and industrial manufacturing more than the asset-light internet businesses of the 2010s.

The shift does not make the companies unattractive by definition. Infrastructure can create durable barriers to entry. A global cloud network with scarce power access, custom silicon, trained personnel, and long-term customer commitments is difficult to replicate. The issue is that returns depend on utilization, pricing, technological obsolescence, and financing costs. The market must value not only software growth but also asset productivity.

Alphabet’s second-quarter 2026 results illustrated the cash-flow pressure. The company generated $39.1 billion of operating cash flow but spent $44.9 billion on property and equipment, producing negative quarterly free cash flow of $5.9 billion. Trailing-12-month free cash flow remained positive, yet the quarter demonstrated how quickly an infrastructure surge can consume the cash generated by an otherwise highly profitable advertising and cloud business.

Amazon showed the same tension in a different form. Its second-quarter operating cash flow increased, AWS growth accelerated, and operating profit reached a new level. Yet trailing-12-month free cash flow turned negative because property and equipment purchases increased sharply. The company’s infrastructure investment primarily reflected AI capacity, and management continued to frame the spending as a response to strong demand rather than speculative overbuilding.

Microsoft occupied the strongest position of the group on July 30 because its operating cash flow expanded enough to preserve substantial free cash flow. It did not spend less in absolute terms. It converted the spending into revenue and cash more visibly. That distinction helps explain why the market’s response was so large.

Capital intensity changes what investors must measure

Traditional software metrics remain useful, but they are no longer sufficient. Revenue growth and operating margin must be examined alongside capital expenditure, lease commitments, depreciation, asset turnover, power costs, and free cash flow. A company can report strong earnings while committing to future infrastructure payments that do not immediately appear in the income statement.

Depreciation policy also becomes more consequential. Extending the useful life of servers or data-center buildings reduces annual depreciation expense, which can improve reported operating income even when the economic asset is aging. Conversely, shortening useful lives can depress earnings while better matching rapid technological change. Investors need to compare accounting assumptions with actual replacement cycles.

Lease classification can alter reported capital expenditure without changing the company’s need for infrastructure. Microsoft’s shift toward operating-lease treatment for more data-center arrangements is a useful example. The economic question is not whether a payment is called capex, rent, or a finance-lease obligation. It is whether the company can earn an adequate return after all cash commitments.

Custom silicon adds another layer. Amazon, Google, Microsoft, Meta, and other hyperscalers are designing or deploying their own processors to reduce dependence on external suppliers, improve performance per watt, and tailor hardware to specific workloads. Successful custom chips can lower cost and strengthen bargaining power. Failed designs can create expensive inventory, software incompatibility, or delayed deployment.

The AI capex cycle therefore moves technology analysis closer to industrial analysis. Capacity planning, supply-chain management, energy procurement, depreciation, and financing become central to the investment case. The market’s July 30 reaction showed that investors are beginning to differentiate companies according to those capabilities.

Are Share Buybacks Really Giving Way to AI Investment?

The Yahoo Finance broadcast framed the change as the end of the buyback era and the beginning of the capital-expenditure era. The direction of travel is credible, but the phrase should not be interpreted literally. Buybacks have not disappeared. Microsoft returned more than $43 billion to shareholders during fiscal 2026, Apple remains one of the largest repurchasers in the market, and many financial, healthcare, and industrial companies continue to reduce their share counts.

What has changed is the competition for cash inside the largest technology companies. A dollar spent on GPUs, data centers, or long-term power contracts cannot be used simultaneously for repurchases. When operating cash flow grows more slowly than infrastructure spending, management must choose among lower buybacks, higher debt, asset sales, reduced acquisitions, or thinner cash reserves.

Historical buyback data shows why the issue matters. S&P 500 companies spent a record $293.5 billion on repurchases in the first quarter of 2025, according to S&P Dow Jones Indices. Spending fell to $234.6 billion in the second quarter and recovered to $249.0 billion in the third quarter. The 12-month total through September 2025 reached a record $1.02 trillion. Repurchases had become a major source of demand for U.S. equities and a mechanism for increasing earnings per share by reducing the number of shares outstanding.

The economic benefit of a buyback depends on price. Repurchasing undervalued stock can transfer value to remaining shareholders. Repurchasing overvalued stock can destroy value even though earnings per share rises mechanically. Buybacks also offset dilution from stock-based compensation, which means the gross amount can overstate the reduction in share count.

AI investment presents a different risk-return profile. It can create new revenue, lower unit costs, and strengthen competitive advantages. It can also produce overcapacity. The correct comparison is not “buybacks good, capex bad” or the reverse. It is the expected return on each use of capital.

Microsoft’s July results favored capex because demand exceeded supply and Azure growth accelerated. The company could argue that another dollar invested in capacity had a high probability of generating revenue. Meta’s results made the tradeoff less comfortable because infrastructure spending absorbed nearly all quarterly operating cash flow. Amazon’s post-close report strengthened its argument by showing a sharp acceleration in AWS and large operating-income growth, but its negative trailing free cash flow confirmed that the funding burden was real.

A lasting decline in buybacks would affect market structure. Repurchases provide persistent demand, reduce public float, and soften the effect of employee equity issuance. If the largest companies cut repurchases while issuing debt or stock to fund AI infrastructure, shareholders would face a less supportive technical environment. The change would be gradual rather than a single break, and it would vary by company.

The more useful question is capital return, not capital category

Investors should compare incremental returns across capital uses. Useful questions include:

• How much additional revenue is associated with each dollar of new capacity?
• How quickly does new infrastructure reach productive utilization?
• What gross margin does the new workload generate after power, depreciation, and networking costs?
• How much of reported demand comes from financially stable customers?
• Are long-term contracts take-or-pay commitments, cancellable reservations, or consumption estimates?
• How much spending is required merely to replace obsolete equipment?
• What portion of buybacks offsets employee dilution rather than reducing the share count?
• Does management have a record of earning returns above the company’s cost of capital?

These questions turn a broad narrative into measurable analysis. Microsoft’s quarter scored well because capacity additions were monetized, Azure growth accelerated, operating cash flow increased, and free cash flow remained positive. The evidence can change in later quarters, which is why the market will continue to scrutinize cash conversion rather than relying on one earnings report.

Amazon’s Results Strengthened the Bullish AI Case After the Close

The Yahoo Finance program ended as investors prepared for Amazon and Apple to report. Amazon’s results, released after the July 30 close, provided the next major test of the AI spending thesis. The company reported second-quarter net sales of $200.6 billion, up 20% year over year, and operating income of $27.5 billion, up 43%.

AWS revenue increased 37% to $42.2 billion, its fastest growth in 18 quarters. AWS operating income reached $16.6 billion, up 64%, and operating margin was approximately 39.4%. Amazon said AWS had reached an annualized revenue run rate of $169 billion. Those figures indicated that cloud demand was accelerating fast enough to support the company’s infrastructure program.

Amazon also disclosed that its AI business and chips business had each exceeded a $25 billion annual revenue run rate. The company highlighted demand for its Trainium processors, Graviton CPUs, Bedrock model platform, and commitments from large AI customers. These claims are company-reported metrics rather than audited segments, so they should be interpreted as management’s operating indicators. Still, they show how Amazon is attempting to monetize AI across hardware, cloud services, and model access.

The result was not uniformly positive. Trailing-12-month free cash flow was negative $7.6 billion, compared with positive $18.2 billion a year earlier. Amazon said the deterioration reflected a $66.1 billion year-over-year increase in purchases of property and equipment, net of proceeds and incentives, primarily for AI investment. The company therefore produced the same mixed picture seen across the sector: excellent operating growth and a much heavier capital burden.

Net income of $62.6 billion also required interpretation because it included $53.4 billion of pre-tax non-operating income, primarily from Amazon’s investment in Anthropic. The gain increased reported earnings per share to $5.75 but did not represent ordinary operating profit from retail, advertising, or AWS. Operating income is a more useful measure for assessing the underlying quarter.

Amazon shares rose in after-hours trading because AWS growth, margins, and total revenue exceeded expectations strongly enough to outweigh concerns about capital spending and the quality of reported net income. The reaction supported the same lesson as Microsoft: investors were not rejecting AI capex categorically. They were demanding evidence that the infrastructure was attached to accelerating, profitable demand.

Apple Offered a Different Test: Strong Earnings, Lower Capex, Weaker Guidance

Apple’s report complicated the simple idea that lower AI spending automatically protects a company from investor concern. Apple posted fiscal third-quarter revenue of $109.4 billion, up 16% year over year, and diluted earnings per share of $2.02, up 29%. The quarter set June-period records for total revenue, earnings per share, iPhone revenue, Mac revenue, and Services revenue.

Company gross margin reached 50.1%, although Apple said tariff refunds added approximately two percentage points. Earnings per share included an estimated $0.11 benefit from those refunds. Removing the unusual benefit would still leave a strong quarter, but the disclosure is important because it affects comparisons with future periods.

Apple’s shares fell in after-hours trading despite the earnings beat. The market focused on management’s September-quarter outlook and warnings about supply constraints in advanced chip production and memory. Revenue growth guidance of roughly 9% to 11% was below the growth rate investors had expected. Services revenue increased but missed some consensus forecasts, weakening one of the company’s highest-margin growth narratives.

Apple represents a different capital-allocation model from the hyperscalers. It does not operate a cloud infrastructure business comparable with Azure, AWS, or Google Cloud. Its AI strategy combines on-device processing, private cloud capacity, software integration, and external partnerships. The lower infrastructure burden supports free cash flow and buybacks, but it also raises questions about whether Apple controls enough compute and model capability to compete as AI features become more central to consumer devices.

The July results showed that strong product demand can coexist with strategic uncertainty. iPhone and Mac growth were impressive, yet investors looked beyond the quarter toward supply, pricing, services momentum, and leadership transition. Apple’s case therefore reinforced a broader lesson: the market was not rewarding spending or austerity in isolation. It was evaluating whether each company’s strategy matched its business model and whether the next quarter could sustain the current growth rate.

Apple’s capital discipline remains a strength. The company can return large amounts of cash to shareholders and avoid some of the depreciation and financing risks confronting cloud operators. Its vulnerability is different: it relies on a concentrated hardware ecosystem, faces global supply-chain constraints, and must prove that AI features can support upgrades and services revenue without undermining margins.

The Semiconductor Rally Reflected Demand, Scarcity, and Relief

The approximately 8% rise in the Philadelphia Semiconductor Index was one of the day’s most important signals. Semiconductor stocks had been under pressure because investors feared that hyperscaler spending would slow, inventory would build, or valuations had moved too far ahead of earnings. Microsoft’s report reduced the probability of an immediate demand collapse.

Cloud companies purchase more than GPUs. AI data centers require CPUs, high-bandwidth memory, networking chips, optical components, power-management systems, storage, cooling equipment, and advanced packaging. A capacity-constrained Azure business therefore supports a broad supply chain. That explains why the rally extended beyond the best-known AI accelerator suppliers.

Scarcity also affects pricing. Qualcomm’s management described high utilization across foundries, memory, testing, and other suppliers. When the entire semiconductor chain is running near capacity, input prices rise. The same AI demand that benefits suppliers can compress margins for companies that cannot pass higher costs to customers immediately.

This creates an uneven cycle. Memory producers and equipment suppliers may gain from shortages, while device-chip companies face cost pressure. Cloud platforms can absorb higher costs if customer demand and pricing are strong. Consumer-electronics companies may struggle if price increases reduce unit demand. The semiconductor rally represented relief about volume, not proof that every participant would earn higher margins.

Technological obsolescence remains another risk. The AI chip market is advancing quickly, and buyers are developing custom processors. A supplier can enjoy strong demand today while losing share when customers shift to internal designs or alternative architectures. The sector’s earnings therefore depend on both the size of the infrastructure cycle and each company’s competitive position within it.

Qualcomm Shows the Opportunity and Risk of Leaving the Smartphone Behind

Qualcomm’s interview during the Yahoo Finance broadcast provided a useful case study in diversification. The company reported fiscal third-quarter 2026 revenue of $9.9 billion, down 4% year over year, GAAP earnings per share of $1.87, and non-GAAP earnings per share of $2.21. Revenue exceeded the midpoint of its guidance, but adjusted earnings fell slightly short of Wall Street expectations and management’s outlook reflected margin pressure.

The core handset business remained weak. QCT handset revenue fell 20% year over year as smartphone demand, customer mix, and Apple’s transition to internal modem technology weighed on results. Apple had historically been a large Qualcomm customer, and management said its participation in the next iPhone generation would be more limited than previously assumed.

Automotive revenue provided the clearest offset. QCT automotive revenue reached approximately $1.6 billion, up 61% year over year, marking another record quarter. IoT revenue increased 9% to roughly $1.8 billion. Combined automotive and IoT revenue rose 28%, showing that Qualcomm’s non-handset businesses were becoming economically meaningful rather than remaining small experimental categories.

The company’s more ambitious claim concerned data centers. Qualcomm said it expected data-center revenue to grow from a small base to about $5 billion in fiscal 2027 and more than $15 billion by fiscal 2029. It expected two global hyperscale customers, one in the United States and one in China, to begin purchasing custom chips, with each relationship expected to generate at least $1 billion of fiscal 2027 revenue.

Those targets are guidance, not achieved results. The ramp depends on production schedules, customer deployment, product performance, software support, and the absence of trade restrictions. Data-center customers have substantial bargaining power and can change architecture plans. Qualcomm must also compete with Nvidia, AMD, Arm-based suppliers, internal hyperscaler designs, and specialized accelerators.

The strategic logic is credible. Qualcomm has decades of experience optimizing performance per watt in mobile devices. Power availability has become a major constraint in AI data centers, making energy efficiency more valuable. Technologies developed for smartphones, PCs, vehicles, and edge devices can be adapted to inference workloads where power and total cost of ownership are critical.

Execution will determine whether the opportunity becomes profitable. Data-center products often require large upfront research and software investment. Early contracts may carry lower margins, and customer concentration can create volatility. A $5 billion revenue target is meaningful, but revenue quality, gross margin, and capital requirements will decide the value created for shareholders.

Qualcomm’s margin problem is immediate

Qualcomm said foundry, memory, testing, and other supplier costs had increased as utilization tightened across the semiconductor chain. The company planned to raise prices, but contracts and product cycles mean cost increases can reach the income statement before price increases reach revenue. That timing gap pressured the near-term outlook.

For its fiscal fourth quarter, Qualcomm forecast revenue of $9.7 billion to $10.5 billion and non-GAAP earnings per share of $2.05 to $2.25. The earnings range was below consensus expectations at the time of the report. The stock declined because investors saw a faster Apple revenue reduction, rising input costs, and uncertainty about how quickly diversification could replace the lost profit.

The distinction between replacing revenue and replacing earnings is crucial. Automotive and data-center sales may have different margins from premium smartphone modems. If $1 of new revenue earns less operating profit than $1 of lost Apple revenue, Qualcomm may need more than a dollar-for-dollar replacement to preserve earnings. Management’s statement that growth in automotive, IoT, and data centers could replace the Apple business is encouraging, but segment profitability will be the more important evidence.

Automotive offers better visibility but slower cycles

Automotive semiconductor programs have long development periods. A chip selected today may not generate production revenue for several years, but once designed into a vehicle platform it can remain in production for an extended cycle. That structure gives suppliers visibility, although delays in vehicle launches or weaker auto demand can shift revenue.

Qualcomm’s automotive business includes connectivity, digital cockpit, driver-assistance, and computing platforms. Automakers increasingly prefer integrated systems that can support software updates and multiple vehicle functions. Qualcomm’s ability to sell several components into one platform can increase revenue per vehicle and deepen customer relationships.

The risk is that automotive customers are demanding on cost, reliability, and long-term support. Gross margins may not match the economics of high-end smartphone chips or licensing revenue. Competition from Nvidia, Mobileye, NXP, Infineon, Renesas, and internal automaker efforts remains significant. The 61% revenue growth is strong evidence of momentum, but not proof that the business will achieve the same returns as Qualcomm’s historical mobile franchise.

Wearables and edge AI remain an option, not a forecast foundation

Qualcomm also described smart glasses, ear-worn devices, necklaces, watches, and other agent-first products as a large potential market. Its Snapdragon platforms already power several wearable categories, and the company has relationships with Meta, Samsung, Google, and other device makers.

The opportunity is plausible because AI assistants may work better when devices can see, hear, and respond continuously. Yet consumer adoption, privacy concerns, battery life, comfort, price, and social acceptance remain unresolved. Management acknowledged that its financial forecast included only a nominal contribution from these devices.

That is the appropriate treatment. Wearables provide upside optionality, but Qualcomm’s near-term thesis should rest on reported automotive growth, measurable IoT demand, contracted data-center programs, and the pace of Apple revenue decline. A market that is increasingly skeptical of distant AI promises will require concrete shipment and margin evidence.

The Federal Reserve and the Bond Market Set a Harder Test for AI Spending

The July 30 equity rally occurred against a macroeconomic backdrop that was less supportive than the stock indexes suggested. One day earlier, the Federal Reserve had held the federal-funds target range at 3.5% to 3.75%. The decision was unusually divided: three voting officials—Beth Hammack, Neel Kashkari, and Lorie Logan—preferred a quarter-point increase. The Federal Open Market Committee said inflation remained above its 2% objective and that uncertainty about the outlook was elevated, including risks associated with Middle East supply disruptions.

A split of that size matters because it changes the market’s understanding of the policy reaction function. Investors had spent much of the previous decade assuming that a weakening economy or falling asset prices would eventually produce easier monetary policy. The July vote showed that a meaningful group of policymakers was more worried about persistent inflation than about providing additional support to growth. That reduces the value of relying on future rate cuts as the solution to expensive financing or ambitious valuation assumptions.

The bond market’s response was more skeptical than the equity market’s. Long-term Treasury yields remained elevated, with the 30-year yield trading around 5.2% and the 10-year yield near 4.7% during the session. Those levels were close to highs not seen since before the global financial crisis. A lower dollar and falling oil prices helped risk assets on the day, but they did not erase the broader increase in the economy’s long-term cost of capital.

For AI investment, this is not an abstract macro issue. A data center is a long-lived asset whose economics depend on years of utilization, power costs, depreciation, maintenance, and technological relevance. When long-term interest rates rise, the present value of distant cash flows declines. Projects that looked compelling under a 3% discount rate may look less attractive when investors demand a return closer to 5%, before adding an equity-risk premium or project-specific risk.

High yields also increase the opportunity cost of cash. A company can earn a meaningful return from short-term securities without taking operating risk. Management teams must therefore show that a new data center, custom chip, or AI application can earn substantially more than the return available on safe assets. The required return is especially demanding when the underlying technology may become obsolete faster than the building that houses it.

Microsoft’s results were powerful partly because they met this tougher standard. The company was not asking investors to accept a distant promise in exchange for current spending. It showed contracted demand, accelerating Azure revenue, strong operating cash flow, and immediate use of new capacity. Those facts do not guarantee a high lifetime return on every dollar invested, but they reduce the risk that the current buildout is entirely speculative.

Meta’s results created the opposite impression. Revenue grew rapidly, but spending grew faster, free cash flow contracted sharply, and operating margin declined. The company may ultimately earn excellent returns from its infrastructure, advertising tools, recommendation systems, and future AI products. At the reporting date, however, investors were being asked to accept more execution risk while the discount rate remained high.

Why the 30-year Treasury yield matters more than the next Fed meeting

Financial markets often focus on the next policy decision, but a one-day change in the federal-funds rate is not the only relevant financing variable. Long-term yields incorporate expectations for inflation, growth, fiscal borrowing, term premium, and the future path of short rates. They affect mortgages, corporate debt, infrastructure finance, equity valuations, and the hurdle rate used in capital budgeting.

A hyperscaler can finance much of its investment internally, which makes it less sensitive to borrowing costs than a leveraged developer. Yet the cost of capital still matters. Shareholders could receive buybacks or dividends instead of funding additional capacity. The company could acquire another business, invest in software, or keep the cash in liquid securities. Choosing a data center means rejecting those alternatives.

Long-term yields also matter to the customers buying AI services. Startups, software companies, and enterprises may reduce experimentation when financing is expensive. A cloud platform can have abundant technical demand but weaker commercial conversion if customers cannot fund pilots, lack confidence in the economic outlook, or struggle to justify an application’s return on investment. The capacity owner therefore faces two related questions: can it build economically, and can its customers monetize what they consume?

The July rally answered the second question more positively for Microsoft than for the market as a whole. Azure’s growth suggested that customers were willing to pay for capacity even under restrictive financial conditions. The narrow breadth of the rally suggested that investors were not extending that conclusion to every company or industry.

Inflation and Consumer Spending Complicated the “Soft Landing” Story

Economic data released around the same time showed an economy that was still expanding but did not provide the Federal Reserve with a clean reason to ease policy. The Bureau of Economic Analysis reported that personal consumption expenditures increased 0.3% in June and real PCE increased 0.4%. Personal income rose 0.2%. Consumers were still spending, but the personal saving rate fell to 2.7%, suggesting that households had less of a cushion than the aggregate spending figures implied.

Inflation remained the larger problem. The PCE price index decreased 0.1% from May because energy prices fell, yet it was 3.7% above its year-earlier level. Excluding food and energy, the index rose 0.1% during the month and 3.3% over 12 months. The monthly readings looked calmer than the annual figures, but both headline and core inflation remained materially above the central bank’s goal.

This combination—resilient spending, low saving, and elevated inflation—supports more than one interpretation. The optimistic view is that households are maintaining demand, employment and incomes are preventing a recession, and inflation may gradually cool as temporary shocks fade. The cautious view is that spending is being supported by a shrinking saving buffer, price levels remain difficult for lower-income households, and the Federal Reserve may need to keep policy restrictive for longer.

Corporate earnings reflected the same division. Lightspeed Commerce, discussed later in the Yahoo Finance broadcast, said spending remained strong among the higher-value retailers and European hospitality businesses it serves. Jersey Mike’s growth story was also linked to consumers willing to pay a premium for a branded meal. Other consumer companies, however, had reported resistance to higher prices, downtrading, and uneven traffic.

The phrase “K-shaped economy” is useful only when it is defined. It describes a situation in which higher-income households and asset owners continue to spend while lower-income consumers experience more pressure from rent, food, transportation, and borrowing costs. Businesses serving affluent customers may therefore report healthy demand even when broader confidence is weak.

That matters for the technology rally because enterprise AI demand is not isolated from the rest of the economy. Advertising budgets depend on corporate revenue. Cloud consumption depends on software activity, e-commerce, media, financial services, and startup funding. Consumer AI devices depend on discretionary income. If spending becomes more concentrated, the largest platforms can continue growing while smaller customers reduce activity.

Microsoft’s diversified customer base provides resilience. It sells to governments, large enterprises, small businesses, developers, schools, and consumers. Amazon has both a cloud platform and a retail network. Meta’s advertising model reaches a broad set of businesses but remains tied to marketing demand. Apple relies more directly on consumer willingness to upgrade premium devices. Qualcomm is exposed to handset units, automotive production, industrial demand, and customer product launches.

The July 30 market response therefore should not be read as a forecast that the entire economy was accelerating. It was evidence that a group of large technology platforms could continue monetizing AI and cloud demand within a mixed economy. Whether that strength spreads depends on labor income, inflation, credit conditions, and the ability of customers to earn a return from new technology.

The Bullish Interpretation: Microsoft Proved the AI Cycle Has Real Revenue

The strongest bullish argument begins with a simple observation: Azure was growing faster at a much larger scale, and demand still exceeded capacity. This is not the pattern of an infrastructure program built solely on internal optimism. Customers were signing commitments, consuming capacity, and contributing to recognized revenue.

Microsoft’s backlog reinforced that conclusion. Commercial remaining performance obligations reached $678 billion. OpenAI accounted for a meaningful part of the increase, which is why the 25% growth rate excluding OpenAI is the more conservative indicator. Even after that adjustment, contracted obligations expanded faster than the company’s overall revenue.

A second bullish point is that AI was improving existing products rather than waiting for an entirely new market to emerge. Microsoft could embed AI in Office, GitHub, security, databases, analytics, customer-service tools, and cloud infrastructure. Each product already had users, distribution, billing, identity, and administrative controls. The company did not need to acquire a new consumer for every AI feature.

Third, the cloud platforms may be building a utility-like layer for digital activity. Enterprises can choose models and applications, but they still need compute, storage, networking, identity, security, and data management. A platform that owns those layers can earn revenue regardless of which model family becomes dominant. Microsoft’s support for multiple models is strategically important because it reduces the need to predict one permanent winner.

Fourth, capacity constraints can protect pricing. When demand exceeds supply, providers have less incentive to discount. Customers with urgent workloads may accept higher prices or longer commitments. Scarcity also reduces the risk that new assets remain idle. This does not guarantee attractive margins, but it improves utilization during the most capital-intensive phase.

Fifth, Microsoft’s financial scale gives it an advantage that is difficult for smaller competitors to reproduce. It can spend tens of billions of dollars per quarter while maintaining investment-grade credit, positive free cash flow, and substantial shareholder distributions. It can negotiate long-term power agreements, reserve semiconductor capacity, finance custom silicon, and build across regions. Those capabilities can create a feedback loop in which scale improves supply, supply improves product availability, and availability attracts more customers.

Amazon’s AWS results supported the same broader thesis. AWS revenue increased 37% to $42.2 billion, its fastest growth in 18 quarters, while AWS operating income rose 64% to $16.6 billion. That performance showed that demand was not limited to one cloud provider. Amazon’s total free cash flow turned negative on a trailing-12-month basis because investment increased dramatically, but the underlying cloud segment generated strong operating profit.

The semiconductor rally also supported the demand argument. If hyperscalers were preparing to reduce spending abruptly, suppliers would be less likely to face tight utilization and input-cost pressure. Qualcomm’s comments about foundry, memory, and testing costs suggested the supply chain was operating at high intensity. Scarcity can produce margin pressure, but it is inconsistent with a near-term collapse in volume.

The market may be moving from “AI excitement” to “AI accounting”

Early stages of a technology cycle are often driven by demonstrations, product announcements, and estimates of total addressable market. The next stage requires measurable economics. Investors begin asking how much revenue is incremental, how much cost is variable, how quickly assets depreciate, and whether customers renew after pilots.

Microsoft’s July report was important because it provided more of those accounting links. It connected capital expenditure with capacity, capacity with Azure growth, growth with operating cash flow, and cash flow with continued shareholder returns. The story was still incomplete—Microsoft did not disclose a separate income statement for generative AI—but it was more tangible than a general claim that AI would transform every industry.

As the cycle matures, companies will be compared on conversion rather than ambition. A company spending $30 billion can outperform one spending $20 billion if the larger program generates much more revenue and profit. A company with slower revenue growth can still create value if it deploys capital selectively and preserves margins. The decisive variable is not absolute spending; it is the spread between the return on investment and the cost of capital.

The Skeptical Interpretation: One Strong Quarter Does Not Settle the Return Question

The bearish or skeptical case does not require believing that AI has no commercial value. It requires believing that current spending, valuations, or expectations assume more value than the industry will ultimately capture. Microsoft’s quarter reduced that risk, but it did not eliminate it.

First, the market’s response itself raised the valuation hurdle. A roughly $450 billion increase in Microsoft’s market value in one day represented a large amount of expected future profit. The company did not add $450 billion of current assets or contracted cash during the session. Investors reduced the discount they applied to the AI strategy and raised their estimates of future earnings. That repricing can reverse if growth slows or margins disappoint.

Second, free cash flow may remain volatile as capital spending rises. Microsoft said fiscal 2027 capital expenditures would increase year over year. Approximately two-thirds of the July-quarter capital expenditure was allocated to short-lived assets such as CPUs and GPUs. Those components may need replacement faster than data-center buildings, and their economic life can shorten when new architectures deliver better performance per dollar or per watt.

Third, depreciation can lag cash spending. The cash outflow occurs when equipment is purchased, but the expense is recognized over its estimated useful life. If companies extend the useful life of buildings or equipment, reported earnings can improve even though the cash has already left the business. Microsoft changed the useful life of certain data-center and office assets from 15 years to 25 years. Management explained that the change reflected operating experience and asset durability, but investors should still separate accounting expense from cash investment.

Fourth, large backlogs can contain concentration risk. OpenAI contributed materially to Microsoft’s remaining performance obligations. A commitment from a strategically important customer can support growth, but it can also create exposure to that customer’s financing, model economics, competitive position, and ability to meet long-term commitments. The 25% backlog growth excluding OpenAI is therefore a more balanced measure than the headline 84% increase.

Fifth, the industry may be building redundant capacity. Microsoft, Amazon, Alphabet, Meta, Oracle, specialized cloud providers, sovereign projects, and large enterprises are all investing. Each participant can rationally see strong demand while the combined industry eventually builds more capacity than customers need. Infrastructure cycles often look tight before supply arrives, then shift rapidly when projects complete at similar times.

Sixth, customer economics remain uncertain. A company can generate revenue by selling expensive AI capacity even if customers are not yet generating attractive returns. During an experimentation phase, enterprises may accept high costs to learn. The more durable test comes at renewal, when chief financial officers compare productivity gains or new revenue with inference, integration, data, security, and staffing costs.

Seventh, competition may transfer value from infrastructure owners to customers. Cloud providers can use custom chips and software optimization to lower costs, but rivals can respond with price cuts. Open-source models can reduce software pricing. Enterprise buyers can split workloads across providers. If compute becomes more efficient faster than usage expands, revenue growth can slow even as the number of AI tasks increases.

Eighth, power remains a physical constraint. Data centers require electricity, transmission, cooling water or alternative cooling systems, land, and grid interconnection. Power availability can delay projects and raise costs. Long-term power agreements may lock in supply but expose companies to regulatory, construction, and commodity risks. The economic advantage may accrue to locations, utilities, or equipment suppliers rather than entirely to the cloud platform.

Ninth, regulation and geopolitics can alter the addressable market. Export controls affect the chips that can be sold into China and other jurisdictions. Data-residency requirements can force regional investment. AI safety, copyright, privacy, and competition rules can increase compliance costs or limit product deployment. A global infrastructure program must be evaluated after those frictions, not on a frictionless total-addressable-market estimate.

Tenth, the rally’s narrow breadth showed that investors were not fully convinced about the economic spillover. A healthy investment boom can support industrial, utility, construction, and service companies, but it can also crowd out other spending and concentrate profit in a few platforms. The weaker equal-weight market suggested that the July session was a company-specific revaluation rather than a universal endorsement of the cycle.

AI Infrastructure Is Not the Dot-Com Bubble, but the Comparison Is Still Useful

Comparisons with the late-1990s telecommunications and internet investment cycle are often made too casually. The companies leading the current AI buildout are profitable, cash-generative, globally scaled, and supported by existing customers. Microsoft, Amazon, Alphabet, and Meta are not pre-revenue startups financed entirely by speculative equity. They own large businesses that can fund infrastructure internally.

The technology is also already commercial. Cloud customers pay for AI training and inference. Developers pay for coding assistants. Advertisers use machine learning to improve targeting and creative production. Consumers subscribe to AI products. Enterprises are deploying agents in service, security, software development, analytics, and internal search.

Those differences weaken a simplistic bubble analogy. Yet the historical comparison remains useful because infrastructure cycles can generate real demand and still produce poor returns for some investors. The internet transformed the economy, but many fiber networks, telecom projects, and technology shares were overfinanced or overvalued. Society benefited even when the original capital providers earned disappointing returns.

The same outcome is possible in AI. Compute may become essential, productivity may improve, and new businesses may emerge, while some data centers, chips, models, and platforms earn less than expected. Technological importance is not the same as investment attractiveness.

A second lesson is that falling unit costs can expand usage while pressuring revenue per unit. Semiconductor performance, model efficiency, quantization, distillation, and software optimization can reduce the cost of inference. Lower prices may unlock new applications, but providers must grow volume faster than price declines to preserve revenue growth.

A third lesson is that bottlenecks change. At one stage, GPUs may be scarce. Later, power or memory may be the constraint. Eventually, customer data, integration talent, or regulation may limit adoption. Investors who value a company solely on the current bottleneck risk missing where economic scarcity moves next.

A fourth lesson is that balance sheets determine survival. Companies with recurring revenue and low leverage can continue investing through a downturn. Highly leveraged projects may be forced to sell assets or refinance at unfavorable rates. Microsoft’s positive free cash flow and large operating cash generation are therefore strategically important even if the direct return on AI spending takes time to mature.

How to Judge Whether AI Capital Expenditure Is Creating Value

Readers do not need access to proprietary data-center models to evaluate the investment cycle. Public companies disclose enough information to build a disciplined framework. The following measures are more useful when considered together than when treated as isolated headlines.

1. Revenue growth in the capacity-consuming business

For Microsoft, Azure growth is a direct indicator. For Amazon, AWS revenue and backlog are central. For Alphabet, Google Cloud growth and profitability matter. For Meta, advertising revenue, engagement, and monetization improvements help show whether infrastructure is improving the core business.

Growth should be compared with capacity additions. If capital expenditure accelerates but cloud revenue slows, the company may be building ahead of demand. That can be rational, but the gap must be explained. If revenue accelerates as new capacity arrives, the market has stronger evidence of utilization.

2. Operating cash flow versus capital expenditure

Operating cash flow indicates how much cash the existing business generates before capital investment. Comparing it with capital expenditure shows whether the company can finance the buildout internally. Microsoft produced $55.4 billion of operating cash flow and spent $41 billion in its July quarter. Alphabet’s quarterly capital expenditure of $44.9 billion exceeded its $39.1 billion of operating cash flow, resulting in negative free cash flow.

One quarter does not establish a permanent trend because working capital and payment timing can cause volatility. A trailing-12-month view is usually more informative. The direction still matters: persistent capital expenditure above operating cash flow increases the likelihood of debt issuance, lower buybacks, or asset financing.

3. Free-cash-flow conversion

Free-cash-flow conversion can be calculated in several ways, including free cash flow as a percentage of revenue or operating income. The purpose is to measure how much accounting profit becomes cash after investment. Declining conversion is not automatically negative during a high-return growth phase, but management should show why the temporary sacrifice will produce future cash.

Microsoft’s positive $19.6 billion quarterly free cash flow was one reason its spending was rewarded. Meta’s $784 million of free cash flow and Amazon’s negative trailing-12-month free cash flow showed a heavier current burden. Those figures do not rank the companies permanently; they identify where investors are assuming more future payoff.

4. Backlog quality, duration, and concentration

Remaining performance obligations and contract backlog can provide visibility, but a headline total is incomplete. Investors should ask how much is expected to convert within 12 months, how much depends on one customer, whether commitments can be canceled, and whether the customer is financially strong.

Microsoft’s disclosure of backlog growth excluding OpenAI was valuable because it helped separate broad commercial demand from one large strategic relationship. Similar adjustments should be sought whenever a single customer, government contract, or related-party agreement materially influences the total.

5. Cloud and segment gross margins

Revenue growth can be less valuable when each additional dollar requires expensive compute. Gross margin helps identify whether pricing and efficiency are keeping pace with infrastructure cost. Microsoft Cloud gross margin of 65% remained high but declined as Azure became a larger part of the mix and AI costs increased.

Segment margins should be interpreted carefully. Cloud providers allocate depreciation, internal software, shared facilities, and stock compensation differently. Comparisons are imperfect. The trend within one company can still reveal whether scaling is improving or diluting economics.

6. Depreciation, useful-life changes, and lease commitments

Capital expenditures create future depreciation. Finance leases and operating leases can move some investment away from the standard cash-capex line while preserving a long-term payment obligation. Investors should therefore examine capital expenditure, assets acquired under leases, depreciation expense, and future lease commitments together.

Useful-life changes deserve special attention. A longer estimated life reduces annual depreciation and lifts reported operating profit, but it does not create cash. The change may be economically justified if buildings last longer, yet fast-moving chips may have a much shorter productive life than the structures around them.

7. Capacity utilization and supply constraints

Management statements that demand exceeds supply are useful when paired with revenue acceleration and evidence that new capacity is monetized quickly. The language becomes less meaningful if it persists while growth slows or if customers are merely reserving capacity they may not use.

Utilization is rarely disclosed directly. Investors can infer it from cloud growth, gross margin, backlog conversion, customer commentary, and whether management describes capacity as a constraint. Pricing changes and the amount of idle equipment can also provide clues.

8. Customer return on investment

The cloud platform’s revenue is the customer’s expense. Sustainable demand requires customers to earn more through productivity, revenue, risk reduction, or cost savings than they spend on AI. Case studies should include measurable outcomes rather than only deployment counts.

Useful evidence includes reduced software-development time, lower call-center cost, faster underwriting, increased conversion, fewer security incidents, or higher employee output. The strongest proof will be renewal and expansion after the initial pilot budget has been reviewed.

9. Share count, buybacks, and debt

Capital spending affects shareholders through more than free cash flow. A company can reduce repurchases, issue debt, or use stock compensation to preserve cash. Investors should track diluted share count, net debt, interest expense, and the relationship between announced repurchase authorizations and actual purchases.

A declining buyback does not automatically signal distress if capital is being redeployed into a higher-return project. A rising share count is more concerning when it reflects compensation or financing without corresponding per-share growth. Per-share free cash flow is therefore more informative than total free cash flow alone.

10. Return on invested capital over a full cycle

Return on invested capital is the final test, but it is difficult to measure during a rapid buildout because the denominator rises before all revenue arrives. Investors can still monitor the direction. Operating profit after tax should eventually grow faster than the capital base if the investment is creating value.

A short-term decline may be acceptable. A permanent decline would suggest that the industry is competing away the benefit or that the assets are earning less than the company’s cost of capital. The relevant horizon is not one quarter, but management should provide milestones that allow the market to test the thesis before the end of a decade.

What Microsoft’s Capital Plan Says About the Next Phase

Microsoft indicated that fiscal 2027 capital expenditure would increase from the already elevated fiscal 2026 level. That guidance means the July report was not the end of the investment cycle. It was a request for investors to accept another year of higher spending based on current evidence of demand.

Approximately two-thirds of the July-quarter capital expenditures were for short-lived assets, primarily CPUs and GPUs. The remaining third related to longer-lived assets such as data-center buildings and infrastructure. This mix helps explain why management expects spending to remain high even after large campuses are constructed: the computing equipment inside them must be expanded and refreshed.

Short-lived assets can produce fast revenue but also create rapid depreciation and obsolescence risk. A new accelerator may deliver much better performance per watt than the previous generation, lowering the economic value of older equipment. Software optimization can extend productive life, and older chips can be reassigned to less demanding workloads, but the pace of innovation makes useful-life assumptions critical.

Long-lived assets create a different risk. Buildings, power connections, and cooling systems can support multiple generations of equipment, which makes them more durable. Their value depends on location, grid access, network connectivity, water or cooling availability, and local regulation. A well-sited data center can remain valuable even if the chips change. A poorly located facility can become constrained by power or permitting.

Microsoft’s scale allows it to manage a portfolio of regions and workloads. Capacity can be allocated between internal products and external Azure customers. The company can also use software to improve fleet efficiency, schedule workloads, and shift demand. These capabilities increase the chance that older assets remain productive.

The risk is that internal demand can obscure external economics. Microsoft can consume infrastructure through its own Copilot products and search services, but that does not automatically establish a market price or profit. Investors should look for growth in paid seats, usage-based revenue, customer expansions, and gross profit rather than relying only on internal deployment counts.

Amazon and Microsoft Now Face the Same Question From Different Starting Points

Microsoft and Amazon both reported accelerating cloud demand, but their financial profiles differ. Microsoft combines cloud infrastructure with high-margin software subscriptions, enterprise applications, and operating systems. Amazon combines AWS with a lower-margin retail and logistics business, advertising, subscriptions, and a rapidly expanding capital program.

AWS generated $16.6 billion of operating income on $42.2 billion of quarterly revenue, an operating margin of approximately 39.4%. That segment remains Amazon’s primary profit engine. The result demonstrated that cloud infrastructure can produce strong current profitability even while Amazon increases spending.

At the consolidated level, Amazon’s trailing-12-month free cash flow was negative $7.6 billion. The company attributed the reversal largely to a $66.1 billion increase in purchases of property and equipment, principally reflecting AI investment. This does not mean AWS was unprofitable. It means the combined company was investing faster than its current operating cash flow could absorb under Amazon’s definition of free cash flow.

Amazon’s retail network can benefit from AI through inventory placement, recommendations, advertising, fulfillment, customer service, and seller tools. Those benefits may not appear as separately identified AI revenue. They can instead improve delivery speed, conversion, labor efficiency, or advertising yield. That makes Amazon’s return analysis more complicated than simply comparing AWS sales with data-center spending.

Microsoft’s enterprise software base gives it more direct opportunities to charge a premium for AI features. Amazon has a stronger position in commerce and logistics, where returns may appear as lower cost or higher retail efficiency. Both companies can create value, but the evidence will look different.

The market’s after-hours reaction to Amazon suggested investors accepted the spending because AWS growth accelerated and operating income increased strongly. That acceptance can be temporary. If the capital program continues to expand while free cash flow remains negative, the market will demand more precise evidence about capacity utilization and future cash recovery.

Meta’s Spending Could Still Work, but the Burden of Proof Increased

Meta’s second-quarter revenue growth of 28% was not weak. Daily active people across its family of applications increased, advertising demand remained healthy, and AI systems likely contributed to recommendations and ad performance. The problem was the relationship between growth and cost.

Total costs and expenses rose 55% to $42.0 billion. Operating income fell 8%, operating margin declined to 31% from 43%, and net income fell 14%. Capital expenditures, including principal payments on finance leases, reached $31.1 billion. Free cash flow fell to $784 million from $8.5 billion a year earlier.

The optimistic case is that Meta is investing before a major increase in monetization. Better recommendation systems can raise engagement, more capable advertising tools can improve conversion, and future personal agents or devices can open new revenue. Meta has a history of investing aggressively and then improving efficiency.

The skeptical case is that competition for talent, chips, power, and models has structurally increased the cost of operating the platform. If every major advertiser and social network can access similar models, some AI benefit may be competed away. Revenue could grow while the incremental margin remains lower than in the pre-AI period.

Meta’s position also differs from a cloud provider because much of its infrastructure supports an advertising business rather than a direct external compute service. It can monetize improved engagement and ad performance, but the link between a dollar of capex and a dollar of revenue is less visible. That opacity increases the importance of margin, free cash flow, and detailed product evidence.

The July comparison with Microsoft therefore should not be reduced to “Microsoft good, Meta bad.” It showed that the market was differentiating between the timing and visibility of returns. Microsoft provided more current evidence. Meta asked investors to underwrite more future benefits.

What the Rally Means for Semiconductor Companies

A sustained hyperscaler investment cycle would support a broad group of semiconductor companies, but the value will not be distributed evenly. Accelerators attract the most attention, yet the data center is a system. Memory bandwidth, networking, optical interconnect, power management, cooling controls, storage, CPUs, and advanced packaging can all become bottlenecks.

Companies positioned at a bottleneck can earn strong margins while supply is tight. The risk is that capacity expansion eventually removes the shortage. Memory and manufacturing equipment businesses are historically cyclical because customers order aggressively during shortages, then reduce spending when inventory and capacity normalize.

Custom silicon introduces another layer. Microsoft, Amazon, Alphabet, Meta, and other hyperscalers are developing or commissioning processors optimized for specific workloads. Custom chips can reduce cost and dependence on external suppliers. They can also expand the market by making inference more economical.

For merchant semiconductor companies, custom silicon is both a threat and an opportunity. It can reduce share in general-purpose accelerators, but it creates demand for design services, intellectual property, networking, memory, packaging, and manufacturing. Qualcomm’s plan to sell customized data-center chips illustrates that shift.

Software remains a strategic barrier. A chip with attractive performance per watt is not enough if developers cannot deploy models easily or if the ecosystem lacks tools, libraries, and support. Nvidia’s software position has historically reinforced its hardware advantage. Competitors must offer either compatibility, superior economics, or a workload-specific benefit large enough to justify migration.

The July semiconductor rally priced in stronger near-term demand, but it did not resolve the competitive structure. Investors should distinguish companies selling scarce components into many platforms from companies dependent on one architecture, one customer, or one stage of the cycle.

Power, Grid Access, and Cooling Are Becoming Financial Variables

AI strategy is increasingly constrained by infrastructure outside the traditional technology sector. A data center can have chips, customers, and financing but still be delayed by transmission capacity, interconnection queues, generation availability, or local permitting.

Power affects both growth and margin. If electricity prices rise, the cost of serving compute increases. A cloud provider may pass some cost to customers, improve efficiency, or shift workloads geographically. Each response has limits. Customers resist price increases, efficiency gains require investment, and data-residency or latency requirements can prevent relocation.

Performance per watt is therefore an economic measure, not merely a technical specification. A chip that produces the same output using less power can allow more revenue within a fixed electrical envelope. It can also reduce cooling requirements and improve total cost of ownership.

Qualcomm emphasized this point in explaining its data-center opportunity. The company’s mobile heritage required efficient computing because battery life and heat are central constraints in phones. Whether that advantage is sufficient in large-scale data centers remains to be proven, but the direction of customer demand supports the strategy.

Cooling creates additional capital and operating costs. High-density racks may require liquid cooling, facility redesign, and new maintenance capabilities. Water availability can become a community and regulatory issue. These constraints influence where capacity is built and how quickly it can be deployed.

Long-term power contracts can reduce price uncertainty but create obligations. Some projects may depend on new natural-gas generation, nuclear power, renewable supply, or storage. Delays can leave equipment waiting for interconnection. The financial model must therefore include more than chip cost and cloud pricing.

The Buyback Debate Needs More Precision

The Yahoo Finance segment described the transition from a “buyback era” to a “capex era.” The phrase captures a real shift: large technology companies are directing more cash toward infrastructure, and aggregate share repurchases may become less dominant. It should not be interpreted as proof that buybacks are ending across corporate America.

S&P 500 companies repurchased a record $293.5 billion of shares in the first quarter of 2025, followed by $234.6 billion in the second quarter and approximately $249.0 billion in the third quarter. Twelve-month buybacks exceeded $1 trillion. Repurchases remained historically large even after quarterly fluctuations.

The more meaningful change is at the company level. A hyperscaler that once used surplus cash primarily for repurchases may now prioritize data centers, chips, power contracts, and model development. The opportunity cost belongs to its shareholders even if total index buybacks remain high.

Buybacks can create value when shares trade below intrinsic value and the company has no better use for the cash. They can destroy value when executed at excessive prices or financed with expensive debt. Capital expenditure can create value when returns exceed the cost of capital and destroy value when assets are underused. Neither category is inherently superior.

Per-share outcomes are the correct focus. Suppose operating profit grows 20% while diluted share count rises 5% and capital requirements double. The shareholder’s result differs from a company that grows profit 15%, reduces share count 3%, and preserves free cash flow. Revenue headlines do not capture that distinction.

Microsoft continued to repurchase shares and pay dividends while investing heavily, which demonstrated financial flexibility. Amazon historically prioritizes reinvestment and does not pay a regular dividend. Meta has added dividends and repurchases but faces a larger current cash burden. Apple remains the most buyback-intensive of the group. Each company’s capital allocation should be judged against its growth opportunities and valuation.

What Happened After the Yahoo Finance Broadcast

The original broadcast occurred during the final hour of regular trading, before Apple and Amazon released results. At that point, Microsoft was already driving the rebound, Amazon shares were rising in anticipation of its report, and Apple shares were lower. The after-hours reports provided a useful test of whether Microsoft’s result was unique.

Amazon strengthened the case that cloud demand remained broad. AWS revenue of $42.2 billion grew 37%, and AWS operating income rose 64%. The company’s total sales increased 20% to $200.6 billion. Investors initially rewarded the acceleration even though the capital program had pushed trailing free cash flow below zero.

Apple delivered strong historical results but a less reassuring forward message. Fiscal third-quarter revenue rose 16% to $109.4 billion, and earnings per share rose 29% to $2.02. The company set June-quarter records in several categories. Shares nevertheless declined after management guided to slower revenue growth and discussed supply constraints.

Together, the reports confirmed that investors were evaluating AI and technology companies through different operating models. Microsoft was rewarded for high spending plus visible monetization. Amazon was rewarded for accelerating AWS growth despite negative consolidated free cash flow. Apple was penalized despite strong current results because the next-quarter outlook and supply picture were less compelling.

The sequence also showed why live-market commentary can become outdated quickly. A stock’s move before earnings reflects expectations, positioning, and uncertainty. The after-hours move reflects the first interpretation of new information, often before analysts have fully reviewed the details. A more reliable conclusion requires separating regular-session prices, after-hours prices, and subsequent trading.

What Investors Should Watch Next

The next phase will be determined by a sequence of operating and macroeconomic data rather than by one earnings day. Several indicators can confirm or challenge the July 30 interpretation.

Employment Cost Index on July 31

The Bureau of Labor Statistics scheduled the second-quarter Employment Cost Index for July 31 at 8:30 a.m. Eastern time. Wage and benefit growth matters to the Federal Reserve because labor costs can influence service inflation. A stronger-than-expected reading would support the officials who preferred a rate increase; a softer reading could reduce pressure on long-term yields.

U.S. employment report on August 7

The July employment report was scheduled for August 7. Payroll growth, unemployment, hours worked, and wage gains will shape the outlook for consumer spending and policy. A very strong report could increase rate expectations, while a sharp weakening could raise concern about enterprise and consumer demand.

Consumer Price Index on August 12

The July CPI report was scheduled for August 12. Investors will look beyond the headline to shelter, services, goods, and energy. A broad acceleration would increase the discount-rate pressure on growth stocks. A sustained cooling trend could support valuations even if the Federal Reserve does not cut immediately.

Personal income and outlays on August 26

The next PCE report was scheduled for August 26. Because the Federal Reserve emphasizes PCE inflation, the release will be central to the policy debate. Consumer spending and the saving rate will also indicate whether demand remains durable or increasingly depends on households drawing down their buffers.

Cloud growth and capex guidance in the next earnings cycle

For Microsoft, the key confirmation would be approximately 45% constant-currency Azure growth, continued capacity constraints, and operating cash flow sufficient to absorb higher investment. Investors should compare the actual results with management’s guidance rather than celebrating a high growth rate in isolation.

For Amazon, the focus will be whether AWS maintains its acceleration and whether consolidated free cash flow begins to recover. For Alphabet, the question is whether cloud profit and advertising cash generation can outgrow capital spending. For Meta, the market will seek evidence that infrastructure investment improves revenue, engagement, or efficiency faster than expenses.

Qualcomm’s data-center ramp

Qualcomm said revenue from custom data-center chips would begin in the December quarter and scale rapidly in fiscal 2027. Investors should watch customer qualification, production timing, gross margin, and the balance between U.S. and China exposure. The company’s $5 billion target will become more credible as shipments, customer concentration, and profitability are disclosed.

Power and semiconductor supply

Foundry utilization, memory pricing, advanced packaging capacity, and data-center power availability will influence both revenue and margin. Persistent shortages can support suppliers but raise costs for customers. Rapid capacity additions can relieve cost pressure while creating the risk of oversupply.

Practical Conclusions for Long-Term Investors

The July 30 rally does not justify buying every AI-linked stock, and the elevated bond yield does not justify avoiding every growth company. The evidence supports a more selective approach.

First, separate infrastructure demand from shareholder return. Data-center construction can remain strong while a particular company earns poor margins or overpays for capacity. Suppliers can benefit from volume even if cloud platforms face lower returns. A growing market does not eliminate company-specific risk.

Second, distinguish cash-rich platforms from leveraged beneficiaries. Microsoft can finance investment internally and maintain shareholder returns. Smaller data-center developers, utilities, component suppliers, and software companies may depend on debt or equity markets. The same interest-rate environment affects them differently.

Third, examine the source of growth. Revenue tied to contracted enterprise workloads is different from revenue tied to a speculative customer or temporary shortage. Advertising gains driven by better conversion are different from consumer subscriptions that depend on novelty. The durability of the customer relationship matters as much as the headline rate.

Fourth, value the company on a range of outcomes. A bullish case might assume sustained cloud acceleration, stable margins, falling unit costs, and broad enterprise adoption. A base case might assume continued growth with lower free-cash-flow conversion. A bearish case might include overcapacity, price competition, slower customer adoption, and a persistently high discount rate.

Fifth, avoid treating one quarter as a permanent regime change. Microsoft’s result was important because it added evidence. It was not a guarantee. The correct response is to update probabilities, then test the thesis against the next quarters.

Sixth, pay attention to per-share results. Market capitalization can rise while intrinsic value per share grows more slowly if dilution, debt, or capital intensity increases. Earnings and cash flow per diluted share provide a clearer measure of what the investor owns.

Seventh, remember that the best company can be a poor investment at the wrong price. Microsoft’s quality, distribution, and cash flow are not in dispute. The investment return depends on the price paid relative to future cash. A 15%-plus daily gain raises the importance of valuation discipline.

Frequently Asked Questions

Why did Microsoft stock rise so much on July 30, 2026?

Microsoft reported faster-than-expected Azure growth, strong cloud revenue, a large contracted backlog, and positive free cash flow despite $41 billion of quarterly capital expenditure. Investors concluded that the company was monetizing new AI capacity faster and more profitably than feared. Positioning after a technology selloff amplified the move.

How much did Microsoft add in market value?

Microsoft added approximately $450 billion in market capitalization during the regular session, based on contemporaneous market reporting. Market capitalization is the share price multiplied by shares outstanding; it is not cash received by the company.

Was the July 30 stock-market rally broad?

No. The major indexes rose strongly because Microsoft, technology, and semiconductor shares have large benchmark weights. Most S&P 500 constituents declined, and the equal-weight index was weaker. The session was better described as a concentrated technology rally than a universal risk-on move.

What was Microsoft’s Azure growth rate?

Azure and other cloud-services revenue increased 43% year over year in Microsoft’s fiscal fourth quarter of 2026. Management forecast approximately 45% growth in constant currency for the following quarter.

How much did Microsoft spend on capital expenditures?

Microsoft reported $41 billion of capital expenditure in the quarter ended June 30, 2026. Approximately two-thirds was associated with short-lived assets such as CPUs and GPUs, while the remainder related to longer-lived data-center infrastructure.

Did Microsoft still generate free cash flow?

Yes. Quarterly cash flow from operations was $55.4 billion, and free cash flow was $19.6 billion. That positive cash generation was central to the market’s favorable reaction because it showed the operating business could absorb a large portion of the investment burden.

Why did Meta fall even though its revenue grew?

Meta’s revenue increased 28%, but costs increased 55%, operating margin fell, and free cash flow declined to $784 million. Investors focused on the immediate cash and margin burden of its infrastructure program and the less direct visibility into when that spending would produce incremental profit.

What did Amazon’s earnings say about AI demand?

Amazon reported 37% AWS growth, the segment’s fastest rate in 18 quarters, and a 64% increase in AWS operating income. The results supported the view that cloud and AI demand was strong across more than one provider. Amazon’s consolidated trailing free cash flow was negative because property and equipment investment increased sharply.

Why did Apple shares fall after strong earnings?

Apple posted record June-quarter revenue and strong earnings growth, but investors focused on a softer September-quarter revenue outlook, supply constraints, and some concern about Services growth. The market was looking beyond the reported quarter to the pace and quality of future growth.

How do high Treasury yields affect AI stocks?

Higher yields increase the discount rate applied to future cash flows and raise the return investors require from capital-intensive projects. They also increase the opportunity cost of using cash for data centers rather than safe securities, dividends, or buybacks. Companies with immediate revenue and cash evidence are better positioned than companies relying on distant profits.

Is big-tech capital spending replacing stock buybacks?

For some hyperscalers, a larger share of cash is being directed toward infrastructure, which can reduce the amount available for repurchases. Buybacks remain historically large across the S&P 500, however. The relevant question is whether each company’s investment earns a higher return than the alternatives and improves per-share value.

What is the biggest risk to the AI infrastructure boom?

The largest combined risk is that industry capacity grows faster than profitable customer demand while interest rates remain high. That scenario could produce price competition, lower utilization, margin pressure, and disappointing returns even if AI usage continues to expand. Other risks include power constraints, chip obsolescence, regulation, customer concentration, and geopolitical restrictions.

Final Assessment

Microsoft’s fiscal fourth-quarter report changed the market’s assessment of the AI investment cycle because it joined three facts that investors had been struggling to reconcile: extraordinary capital expenditure, accelerating cloud demand, and positive free cash flow. The company showed that large-scale AI investment can produce current revenue and cash rather than only a distant strategic promise.

The market’s response was rational in direction, even if the magnitude reflected positioning and valuation expansion. Microsoft reduced uncertainty around Azure capacity, customer demand, and near-term growth. Amazon then provided independent evidence that cloud demand was accelerating. The semiconductor rally showed that investors expected the infrastructure cycle to remain strong.

The cautionary evidence was equally important. The rally was narrow. Long-term Treasury yields remained high. Inflation was above target. Meta’s cash generation deteriorated as spending increased. Amazon’s consolidated free cash flow turned negative. Apple’s strong quarter did not prevent a negative after-hours response when guidance disappointed. Qualcomm’s diversification targets remained dependent on execution and margin.

The correct conclusion is therefore not that AI capital expenditure has been vindicated without qualification. It is that the market has begun separating productive spending from spending that still requires proof. Microsoft moved into the first category on July 30 because its revenue, backlog, operating cash flow, and capacity commentary aligned. Other companies will have to show their own chain from capital to utilization, from utilization to customer value, and from customer value to per-share cash flow.

That chain is the central financial story of the AI era. The companies able to demonstrate it can continue receiving capital even in a high-rate environment. Those that cannot may discover that technological ambition is not enough when investors can earn more than 5% in long-term government bonds and demand a premium for taking execution risk.

Sources

• Reuters: U.S. stocks rally as Microsoft offsets Federal Reserve concerns, July 30, 2026
• Microsoft: Fiscal 2026 fourth-quarter earnings call and financial commentary
• Meta Platforms: Second-quarter 2026 results
• Amazon: Second-quarter 2026 results
• Apple: Fiscal third-quarter 2026 results
• Reuters: Apple results and September-quarter outlook, July 30, 2026
• Qualcomm: Fiscal third-quarter 2026 results
• Qualcomm: Data-center and diversification strategy
• Reuters: Qualcomm cost pressure and Apple transition, July 30, 2026
• Federal Reserve: FOMC statement, July 29, 2026
• U.S. Bureau of Economic Analysis: Personal income and outlays, June 2026
• Alphabet: Second-quarter 2026 earnings release
• S&P Dow Jones Indices: S&P 500 first-quarter 2025 buybacks
• S&P Dow Jones Indices: S&P 500 second-quarter 2025 buybacks
• S&P Dow Jones Indices: S&P 500 third-quarter 2025 buybacks
• U.S. Bureau of Labor Statistics: 2026 economic-release calendar

Affiliate disclosure: Businessfinance.news may earn compensation from qualifying actions completed through selected links on this website, at no additional cost to the reader. Affiliate relationships do not influence our editorial reporting, analysis, or conclusions.