Recommended Free Tools
Microsoft and Meta are both committing extraordinary sums to AI, but their July 29, 2026 earnings reports drew opposite reactions: Microsoft shares rose about 2.4% after hours, while Meta fell about 6.2%. Investors were not simply voting for or against AI. They were judging how clearly each company could connect its spending to demand, revenue and eventual cash returns.
What happened in Microsoft’s and Meta’s latest earnings?
Microsoft’s fiscal fourth-quarter report showed approximately $90 billion in revenue, strong Azure growth and continued demand for AI infrastructure. Its shares initially rose after the release. Meta’s second-quarter revenue grew 28%, but expenses rose 55% to about $42 billion, and the company raised its 2026 capital-expenditure outlook. Its shares fell in after-hours trading. Those figures, and the after-hours moves, were reported by Axios; the revenue figure for Microsoft was reported as approximate by The Associated Press.
That contrast makes “earnings soar” an inaccurate description of both reports. Microsoft’s business and outlook received a positive response; Meta’s revenue grew, but investors focused on the rising expense burden and uncertain payback. The reports were released July 29, 2026, according to Microsoft’s earnings announcement.
Microsoft: strong cloud demand, with a costly capacity buildout
Azure growth and AI workloads were central to the favorable reading of Microsoft’s results. Microsoft has said demand for AI capacity exceeds what it can currently supply. That is a management statement about demand, not proof that every planned data center or accelerator will earn an attractive return. Commercial commitments and backlog can improve visibility, but they may be fulfilled over time and do not equal current recognized revenue or cash flow.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsMicrosoft’s calendar-2026 capital-expenditure plan is roughly $190 billion. Management attributed about $25 billion of that amount to higher component prices. It also said roughly two-thirds of quarterly capex went to short-lived assets, primarily GPUs and CPUs, with the rest going to longer-lived infrastructure. These are management disclosures in its FY2026 Q3 earnings-call materials. The calendar-year plan should not be confused with a fiscal-year total.
Microsoft can monetize AI through several routes: Azure compute sold to customers, enterprise software subscriptions, security and developer services, and products such as Microsoft 365 Copilot. These channels make the revenue path more direct than an AI assistant that has yet to establish a standalone business. But the available figures here do not quantify paid Copilot seats, retention or revenue, so adoption should not be treated as a substitute for reported monetization.
Meta: growing advertising revenue, faster-rising costs
Meta’s 28% revenue growth shows that the top line was expanding, but its 55% expense growth and approximately $42 billion expense total point to a widening cost burden. Meta raised its 2026 capex range to $130 billion–$145 billion from its earlier $125 billion–$145 billion range. The updated range, issued July 29, is tied primarily to AI infrastructure, data centers and building advanced AI capabilities; see Meta’s Q2 2026 earnings-call materials. The earlier range appears in its SEC filing.
Meta’s AI investment can create value without a separate line item called “AI revenue.” Better ranking and recommendations can increase engagement, improve ad relevance and support advertising performance. That indirect contribution is economically meaningful, but it is different from demonstrating that Meta AI, AI agents or enterprise services have become large, profitable standalone businesses. The company’s filings identify AI initiatives, infrastructure commitments, competition, regulation and advertising dependence among risks to future results.
Meta is also investing in talent and Meta Superintelligence Labs as well as physical capacity. Infrastructure spending is capex; research, compensation and other operating costs affect expenses differently. The distinction matters: a company can report accounting profit while heavy capex reduces free cash flow, because capex is not deducted from operating income all at once.
Why did the stocks react differently?
Markets respond to results relative to expectations and to guidance about what comes next, not just whether revenue grew or a prior estimate was exceeded. The July 29 after-hours moves—Microsoft up about 2.4% and Meta down about 6.2%—are one session’s response, not a definitive verdict on either company’s business.
- Microsoft received credit for a clearer near-term revenue channel. Azure consumption, commercial commitments and enterprise software give investors observable places to look for AI-related demand. Management’s claim that capacity is constrained supports the case that new infrastructure can be sold, although it does not establish the eventual margin or return.
- Meta faced a harder spending question. Revenue was growing, but expenses were growing faster and the capex range was enormous. Investors have less direct evidence of standalone AI revenue, even though AI may already be improving advertising and recommendations.
- Expectations can reverse the apparent meaning of “good” news. A stock can fall after strong revenue if spending or forward guidance disappoints; it can rise after huge spending if investors believe monetization is arriving quickly enough.
The market reaction was therefore about the expected price, timing and return on investment—not necessarily a rejection of AI itself.
How large is the Big Tech AI spending cycle?
A broader market estimate put 2026 capital spending by Alphabet, Amazon, Meta and Microsoft at as much as $720 billion, primarily for AI data centers, according to AP coverage of the investment cycle. That is a reported estimate, not one universally standardized accounting total.
Cross-company capex comparisons are imperfect. Companies use different fiscal calendars and accounting presentations; totals may differ in their treatment of finance leases, leased capacity, land, buildings, networking, energy systems and other infrastructure. Some investment is expensed rather than capitalized, and assets have different useful lives. Microsoft’s disclosure that much of its quarterly capex went to short-lived GPUs and CPUs illustrates why a single annual capex number does not capture the full economic cost of operating and replacing AI infrastructure.
Rank #4
What would make the AI buildout a bubble?
A bubble is not simply a fast-growing technology or a large investment. In this context, the concern is that spending and valuations could become detached from durable customer demand and profitable returns. The possibility of overbuilding data-center capacity is part of the debate described in AP’s coverage; it is a risk thesis, not an established conclusion.
Warning signs to test
- Infrastructure capacity grows faster than paying demand, leaving expensive equipment idle.
- Customers experiment with AI but do not renew, expand usage or pay enough to cover compute and support costs.
- Revenue from AI-enabled products rises, but margins and free cash flow weaken as depreciation, energy, cooling and replacement costs accumulate.
- Companies keep raising spending while pushing return targets further into the future.
- Economics depend on a small circle of vendors and customers repeatedly financing one another’s growth, rather than on broad end-user value.
- Stock valuations require years of rapid growth and leave little room for slower adoption, lower prices or execution setbacks.
Why “all demand is fake” is too simple
- Microsoft says AI demand exceeds available capacity, and cloud providers can sell compute to outside customers.
- AI is already being applied to advertising, recommendations, software development, search and enterprise productivity, though the degree of incremental profit varies.
- Microsoft and Meta have substantial existing businesses and cash-generating operations, unlike ventures that depend entirely on outside financing.
Real demand and overbuilding can coexist. Demand may be genuine while competitors collectively build too much capacity, prices fall, or chip economics deteriorate faster than expected. A further complication is that current capacity constraints can obscure what utilization and pricing will look like once supply catches up.
Microsoft and Meta monetize AI differently
| Dimension | Microsoft | Meta |
|---|---|---|
| Most visible route to value | Azure compute and enterprise software, alongside security, developer tools and Copilot products. | AI-assisted advertising, ranking, recommendations and engagement across its consumer platforms. |
| What investors can measure | Cloud growth, AI-related consumption, commercial commitments and eventually paid product adoption. | Ad impressions, price per ad, engagement and conversion; direct AI-product revenue is less established in the available figures. |
| Main spending question | Will infrastructure produce enough incremental cloud and software revenue to offset capex, depreciation and replacement costs? | Will advertising improvements and future AI products justify infrastructure and talent costs without eroding cash generation? |
| Important risk | Capacity constraints can delay sales, customers may optimize usage, and short-lived hardware can lose economic value quickly. | Advertising remains central while direct AI monetization is less mature; heavy spending recalls investor concerns about the earlier metaverse cycle, though the analogy does not prove the outcome will be the same. |
Microsoft’s advantage is that AI capacity can be sold through an established cloud business as well as used to strengthen software products. Its risk is that demand, backlog and revenue do not guarantee attractive margins or free cash flow after the infrastructure bill. Meta’s advantage is distribution: it can place AI features across Facebook, Instagram, WhatsApp and Messenger, while AI may improve its core advertising system. Its risk is that much of the benefit remains indirect while infrastructure and talent costs are immediate.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Accounting details that can change the picture
Capex is not the same as expense or cash generation
Capex requires cash when infrastructure is acquired or built, but it generally affects reported operating income over time through depreciation. A rapidly expanding data-center fleet can therefore coexist with strong reported operating income while free cash flow is under pressure. Conversely, depreciation schedules may spread the accounting cost over years even if accelerators become less useful sooner. Investors should examine asset useful lives, depreciation and replacement needs alongside the headline capex figure.
Finance leases and capacity commitments also matter. They can create future payment obligations or secure infrastructure without appearing in a simple cash-capex comparison in the same way as owned equipment. Company disclosures should be read for lease payments, purchase commitments and the timing of obligations rather than treating capex guidance as the whole funding requirement.
Separate Microsoft’s operating results from OpenAI investment accounting
Microsoft’s reported GAAP earnings have at times been materially affected by accounting gains or losses associated with its OpenAI investment. Those items are not Azure sales, Copilot subscriptions or operating AI revenue. Microsoft’s FY2026 Q2 release disclosed a material OpenAI-related effect on GAAP earnings, and its FY2026 Q3 release separately reported the investment’s effect. Readers should compare GAAP results with the company’s adjusted presentation and operating performance, while keeping investment-accounting effects distinct from customer monetization.
What to watch in the next earnings reports
No single metric can settle the return question. Read revenue, profitability and investment together, and distinguish reported AI revenue from business performance that AI may be helping indirectly.
Demand and monetization
- Azure growth and any disclosed AI-related cloud consumption.
- Paid Copilot seats, retention and expansion, rather than adoption claims alone.
- Meta ad impressions, price per ad, engagement and conversion, which can reveal whether AI is strengthening the advertising engine.
- Commercial bookings and remaining performance obligations, interpreted in light of how long they take to become revenue.
- Any separately disclosed external AI-cloud revenue or direct AI-product monetization from Meta.
Profitability and cash
- Gross and operating margins alongside AI revenue or usage growth.
- Free cash flow after capex, not just accounting earnings.
- Depreciation and amortization, infrastructure utilization and the useful lives assigned to accelerators.
- Whether operating income can grow while the cash cost of capacity continues to rise.
Capital intensity and returns
- Quarterly capex and updated annual guidance, keeping calendar and fiscal years separate.
- Capex as a share of revenue, and the mix of short-lived equipment versus longer-lived infrastructure.
- Finance leases, purchase commitments and costs for power, land, networking and cooling.
- Whether capex growth eventually slows while AI-related revenue and operating income continue to expand.
- Customer renewals and expansion, which help distinguish lasting use from pilots.
For Microsoft, the key edge case is strong AI or Azure growth accompanied by falling free cash flow: the question is whether this is a temporary timing gap or a structural return problem. For Meta, rising capex alongside operating-income growth would not by itself settle the issue; the cash cost and duration of the investment still matter. In both cases, customer self-hosting, multi-cloud purchasing, rapid efficiency gains, power constraints or delayed construction can alter the economics in either direction.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
