Meta’s Profit Falls 14% as AI Spending Accelerates

Meta landed across Google News after reporting a 14% profit decline, despite another quarter of rapid revenue growth. The company earned $15.85 billion during the second quarter of 2026, down from $18.34 billion one year earlier. Revenue rose 28% to $60.80 billion.

Those figures create a sharper conflict than the headline alone suggests. Meta says artificial intelligence is improving advertising, recommendations, product development, and user engagement. However, the company is spending faster than those benefits are reaching its bottom line.

Microsoft provided an immediate comparison. It also increased AI infrastructure spending, but its quarterly profit grew rather than fell. That contrast turned Meta’s results into a test of whether its AI strategy can generate returns before costs reshape the business.

What Meta’s Earnings Actually Changed

Meta’s advertising engine kept growing, but it no longer concealed the financial weight of the company’s AI expansion.

Meta reported its second-quarter results on July 29, covering the three months ending June 30. Its quarterly results showed revenue increasing from $47.52 billion to $60.80 billion.

Advertising remained the foundation of that performance. Advertising revenue reached $59.36 billion, compared with $46.56 billion one year earlier. Ad impressions across Meta’s applications increased 14%, while the average price per advertisement rose 12%.

Those numbers suggest that Facebook and Instagram still have considerable pricing power. They also indicate that Meta’s recommendation and advertising systems are producing measurable commercial gains.

Costs moved even faster. Total costs and expenses increased 55%, reaching $42.03 billion. Operating income fell 8% to $18.78 billion, while the operating margin narrowed from 43% to 31%.

Net income declined to $15.85 billion. Diluted earnings per share fell 13%, from $7.14 to $6.18. Analysts surveyed by FactSet had expected $7.19 per share, although revenue exceeded their average forecast.

Not all of the expense increase came directly from building AI systems. Meta recorded $2.40 billion in charges related to legal proceedings. It also recognized $1.18 billion in severance expenses connected with layoffs announced in May.

That distinction matters. It would be inaccurate to describe the entire profit decline as an AI infrastructure bill. Legal charges and severance affected the quarter, while a higher tax rate also reduced net income.

However, those items do not remove the broader spending pressure. Research and development expenses rose from $12.94 billion to $21.66 billion. Depreciation and amortization increased from $4.34 billion to $6.36 billion.

Capital expenditures reached $31.08 billion during the quarter. That measure includes infrastructure purchases and principal payments on finance leases. Much of Meta’s AI investment appears there before depreciation gradually reaches future income statements.

Meta also raised the lower end of its full-year capital expenditure forecast. It now expects spending between $130 billion and $145 billion, compared with its previous range of $125 billion to $145 billion.

The company expects total 2026 expenses between $165 billion and $169 billion. The higher lower bound incorporates the quarter’s legal charges, but infrastructure remains central to the long-term increase.

These results changed the market’s question. Investors are no longer asking whether Meta can generate strong advertising growth. They are asking whether that growth can outrun the expanding cost of compute, talent, and depreciation.

Why Google News Focused on Cash Flow

The most revealing figure was not the 14% profit decline. It was the collapse in cash left after Meta funded operations and infrastructure.

Meta generated $31.86 billion in operating cash flow, up from $25.56 billion one year earlier. That result shows the underlying business still produces an extraordinary amount of cash.

Yet free cash flow fell 91%, from $8.55 billion to $784 million. Meta defines free cash flow as operating cash flow minus property purchases and finance lease principal payments.

This metric is not identical to profit. Capital expenditures do not normally reduce net income immediately. Instead, their cost reaches the income statement over time through depreciation.

That accounting difference explains why cash flow deserves attention. Meta can report billions in quarterly profit while spending nearly all its operating cash on servers, data centers, networking equipment, and related infrastructure.

The company purchased $30.12 billion of property and equipment during the quarter. It made another $962 million in principal payments on finance leases. Those two figures nearly consumed its entire operating cash flow.

Free cash flow does not represent cash freely available for every corporate purpose. Meta makes that limitation explicit in its financial disclosure. Even so, the year-over-year decline captures the scale of its current investment cycle.

Meta’s balance sheet is changing alongside that spending. Long-term debt increased from $58.74 billion at the end of 2025 to $83.66 billion on June 30. Cash and marketable securities totaled $90.26 billion.

The company also reported no share repurchases during the first half of 2026. It had spent $22.92 billion on repurchases during the comparable period of 2025.

None of these figures suggests an immediate financing crisis. Meta remains highly profitable, holds substantial liquid assets, and operates one of the world’s largest advertising businesses.

The issue is capital allocation. Every dollar committed to infrastructure becomes a dollar unavailable for repurchases, acquisitions, dividends, or another investment during that period.

Meta’s spending also carries a delayed burden. Servers and accelerators lose value as they age, while newer processors can deliver better performance per unit of electricity. Today’s purchases therefore create future depreciation and replacement needs.

Management must maintain enough infrastructure to run existing products, train new models, and support future services. It must also avoid building capacity that customers or internal products do not use efficiently.

This tension reaches beyond investors. Product teams depend on access to compute, while advertisers depend on measurable performance improvements. Users experience the result through recommendations, creative tools, assistants, and generated content.

For knowledge workers, the flood of claims and financial disclosures creates another problem. A structured personal knowledge system can preserve original figures before headlines reduce them to a single percentage.

The quarter therefore produced two valid readings. Meta’s core operations generated more cash, but its infrastructure program absorbed almost all of it. Both statements must remain visible.

That is why the earnings story spread so widely through Google News. The headline number was profit, but the deeper argument concerned how long Meta can invest at this intensity.

Meta Says AI Is Already Paying Off

Meta’s defense rests on current advertising gains, not only a distant promise of general-purpose AI products.

Mark Zuckerberg told investors that AI was accelerating every major part of Meta’s core business. He pointed to content recommendations, advertisement ranking, creative tools, and faster product development.

The argument starts with recommendations. Meta is integrating large language models into systems that decide which posts, videos, and advertisements people see. A large language model predicts and generates language after learning patterns from extensive data.

Zuckerberg said these models give recommendation systems a deeper understanding of content and user interests. Better recommendations can increase time spent in an application, creating more opportunities to show advertisements.

The same logic applies to advertising. Meta uses AI to predict which advertisement is relevant to a particular user. It also helps advertisers generate images, videos, copy, and variations for different audiences.

During the earnings call, Zuckerberg said nine million small businesses were using at least one of Meta’s AI advertising tools.

That figure comes from Meta and has not been independently audited as a measure of business value. It shows distribution, but it does not reveal how frequently each business uses the tools.

Still, the advertising results provide supporting evidence. Ad impressions increased 14%, average prices rose 12%, and advertising revenue grew 27%. Those gains occurred while Meta continued expanding AI across its applications.

Meta’s daily audience also grew. An average of 3.60 billion people used at least one of its applications each day during June, up 3% from a year earlier.

Zuckerberg said Instagram reached two billion daily users. Threads passed 500 million monthly users, according to the company. These platforms give Meta an enormous channel for distributing new AI functions.

Distribution has always been one of Meta’s advantages. It can place an assistant or generated-media feature inside applications that people already use. A startup must often acquire those users individually.

Meta is also pursuing revenue beyond advertising. Zuckerberg described opportunities involving application programming interfaces, business agents, compute sales, and services for large customers.

An application programming interface lets software products exchange requests and data through defined rules. Business agents are AI systems designed to complete tasks, answer customers, or assist employees with limited human direction.

These opportunities remain less proven than the advertising gains. Meta has not disclosed enough revenue detail to show that enterprise AI has become a material business line.

The company also launched Meta One, a subscription offering that combines tools and AI features across its applications. Management said it expects to introduce different options as demand develops.

That creates a two-stage investment case. AI must first improve the mature advertising business. It must then establish new products capable of supporting the infrastructure built for future demand.

The first stage has supporting evidence. The second remains largely a management forecast.

Meta’s advertising results cannot prove that every AI project deserves funding. Recommendation improvements and creative tools might produce returns while frontier-model research or enterprise services deliver less.

The company does not disclose project-level returns. Investors therefore see aggregate revenue growth beside aggregate spending, without a clear map connecting each investment to income.

This reporting gap makes management credibility important. Zuckerberg argues that the results are already visible. Skeptics want a more direct relationship between AI spending, incremental revenue, and durable margins.

Meta must eventually translate internal performance claims into financial disclosure. Otherwise, strong advertising growth will remain open to competing explanations, including pricing, engagement, currency movements, and broader market demand.

Microsoft Exposed Meta’s Hardest Comparison

The central conflict is not Meta against another social network. It is Meta’s spending model against Microsoft’s clearer AI monetization path.

Microsoft reported major infrastructure spending during the same earnings period. Its capital expenditures increased sharply as it expanded cloud and AI capacity.

However, Microsoft also reported a 31% increase in net income to $35.8 billion. Its performance contrasted with Meta’s 14% decline and helped produce a split market reaction.

An earnings comparison reported that Meta shares fell in after-hours trading while Microsoft shares rose. Both companies were spending heavily, but investors interpreted their returns differently.

Microsoft can connect AI investment directly to Azure, its cloud platform. Customers purchase computing capacity, storage, databases, and AI services through established enterprise contracts.

Meta lacks an equivalent cloud business at comparable scale. Its immediate returns depend mainly on making advertisements and recommendations more effective.

That does not make Meta’s strategy inferior. Improving a large advertising platform can create substantial revenue. The problem is measurement and timing.

Microsoft can report cloud revenue growth alongside infrastructure spending. Meta combines the benefits of AI with many other forces affecting advertisements, including engagement, auction prices, and advertiser demand.

Meta’s 28% revenue growth was faster than Microsoft’s total corporate growth during the comparable reporting period. Yet Meta’s operating income declined because expenses rose even faster.

The difference becomes clearer at the margin level. Meta’s operating margin fell 12 percentage points, from 43% to 31%. The business generated more revenue but retained less of each dollar as operating profit.

Microsoft’s approach has risks of its own. Much of its capital spending supports processors and graphics chips that require replacement. Cloud companies can face margin pressure when capacity costs rise faster than customer usage.

Meta is exploring compute sales, which would move it closer to that model. According to Zuckerberg, the company sees an opportunity to provide computing services directly to large customers.

Entering that market would create potential revenue from infrastructure that might otherwise sit underused. It would also put Meta into competition with Amazon Web Services, Microsoft Azure, Google Cloud, and specialized providers.

Those businesses require more than spare processors. Customers expect service guarantees, security controls, developer tools, technical support, regional availability, and predictable performance.

Meta has experience operating global infrastructure for its own applications. Selling reliable enterprise capacity is a different commercial and operational challenge.

The wider industry is already confronting the same problem. A cash flow analysis found growing pressure across Alphabet, Amazon, Microsoft, and Meta as capital requirements increased.

Alphabet’s cloud growth gives it a clearer route from infrastructure to customer revenue. Amazon and Microsoft can also sell capacity through mature cloud platforms.

Meta is attempting a broader route. It wants AI to improve advertising, generate consumer experiences, support wearable devices, create enterprise agents, and potentially become a compute service.

That breadth can produce several revenue sources. It can also spread investment across projects with different schedules and uncertain economics.

The Microsoft comparison therefore matters more than a conventional Meta-versus-Google rivalry. It asks whether investors will accept large spending without a separately reported AI revenue line.

Meta’s response is that AI already strengthens its core business. Microsoft’s results show why that answer might not satisfy the market indefinitely.

Investors want profitable growth while infrastructure expands. Microsoft delivered both during the quarter. Meta delivered rapid revenue growth, falling profit, and a much smaller free cash flow balance.

The Profit Drop Does Not Prove AI Failed

Meta’s quarter weakens an easy success narrative, but it does not establish that AI spending caused every dollar of lost profit.

The 14% decline attracts attention because it compresses the story into one number. That number needs context.

Legal charges and severance totaled $3.58 billion. Meta also reported a higher effective tax rate, which increased from 11% to 16%.

Operating income fell by $1.67 billion from the previous year. Net income declined by $2.49 billion. Those movements cannot be attributed solely to servers, models, or AI employees.

A balanced profit report noted both the legal expenses and severance costs. It also highlighted Meta’s rising infrastructure commitments and weaker free cash flow.

Meta expects operating income for 2026 to exceed its 2025 result. That forecast indicates management views the second-quarter pressure as compatible with full-year profit growth.

The company also projected third-quarter revenue between $61 billion and $64 billion. The midpoint was below the analyst consensus cited after the results, adding uncertainty to the near-term outlook.

Forecasts are not results. Advertising demand can shift with economic conditions, consumer behavior, currency movements, regulation, and competition.

Meta also faces legal and reputational risks unrelated to AI. The company warned that youth-related lawsuits and regulatory scrutiny might produce material losses.

Those issues complicate the investment narrative. AI improvements could strengthen advertising while litigation creates new expenses. A single profit figure cannot isolate those effects.

The opposite mistake is equally risky. Strong advertising growth does not prove that every infrastructure project will earn an acceptable return.

Meta’s research and development expenses increased by $8.71 billion year over year. Management has not provided a detailed allocation showing how much went to AI, mixed reality, content systems, or other programs.

Reality Labs produced $431 million in revenue and an operating loss of $4.62 billion. Its quarterly loss was slightly larger than one year earlier.

Some infrastructure and research can support both AI and hardware projects. Without more granular disclosure, readers should avoid treating every expense category as a clean AI measurement.

Meta’s claim that AI is improving advertisement performance is credible but incomplete. Advertisers care about conversion rates, acquisition costs, brand safety, creative quality, and incremental sales.

The company has not disclosed enough standardized data for outsiders to compare those outcomes across quarters. Usage figures for creative tools reveal adoption, not necessarily profitability.

Generated content presents another uncertainty. More personalized material can increase engagement, but excessive synthetic content might reduce trust or overwhelm users.

Recommendation systems also shape what people see. Improvements measured through engagement do not automatically address concerns about mental health, polarization, misinformation, or addictive product design.

Minda Smiley of Emarketer told the Associated Press that Meta’s optimistic AI message contrasts with negative sentiment around social media harms. That credibility problem can affect how users and regulators interpret new products.

Meta must therefore succeed across three dimensions. It needs technical performance, commercial returns, and public acceptance.

The quarter supplied evidence for the first two, but the evidence remains uneven. Advertising is growing, AI tools are reaching businesses, and infrastructure spending is rising much faster.

It supplied less evidence for durable enterprise revenue or consumer willingness to pay. It also offered no final answer about the long-term margin created by the investment program.

The correct reading is narrower than either extreme. Meta’s AI strategy has not failed, but its financial burden has become too large to treat as a distant side project.

Three Signals to Watch After the Google News Cycle

Meta’s next test will come from margins, new AI revenue, and whether infrastructure spending keeps consuming operating cash.

The first signal is the third-quarter operating margin. Meta has already guided revenue between $61 billion and $64 billion, giving investors a range against which to measure expenses.

A stable or improving margin would support management’s argument that the second quarter contained unusual legal and severance costs. Another sharp decline would strengthen concerns about persistent infrastructure and depreciation pressure.

The composition of expenses will matter as much as the total. Legal charges can disappear quickly, while depreciation from new data centers can remain for years.

Investors should compare revenue growth with research spending, depreciation, and cost of revenue. If those costs repeatedly grow faster, AI gains are not yet preserving Meta’s former economics.

The second signal is disclosed revenue from AI subscriptions, enterprise agents, APIs, or compute services. Meta has described each opportunity, but it has not separated them as meaningful financial categories.

A material disclosure would strengthen the case that Meta is building a second commercial engine. Continued silence would leave the company dependent on indirect benefits within advertising.

Meta One offers one possible indicator. Subscriber growth, retention, and usage would reveal whether consumers or creators will pay directly for AI features across Meta’s applications.

Enterprise agents present another test. Business customers need reliable systems that can handle support, sales, and internal workflows without creating unacceptable errors or security problems.

Compute sales would offer the clearest infrastructure link. However, Meta would need to show that external demand improves utilization without distracting from its consumer platforms.

The third signal is free cash flow. The second-quarter figure of $784 million represented the most visible consequence of Meta’s spending program.

One quarter does not establish a permanent trend. Infrastructure purchases can arrive unevenly, and major projects can make quarterly comparisons volatile.

Still, repeated low free cash flow would change Meta’s flexibility. It could limit repurchases, increase borrowing, or force management to choose among competing projects.

Capital expenditures will remain central. Meta’s current 2026 forecast spans $130 billion to $145 billion, leaving a considerable difference between its lower and upper bounds.

Movement toward the upper end would show that demand for computing capacity remains intense. Movement toward the lower end might signal improved efficiency, slower construction, or reduced requirements.

Readers should also compare Meta with Microsoft, Alphabet, and Amazon. If their cloud businesses turn spending into visible revenue faster, Meta’s indirect advertising model will face more scrutiny.

If Meta restores margins while expanding AI products, the second-quarter decline will look like a costly transition. If costs keep outpacing revenue, the quarter will look like an early warning.

That is the question worth carrying beyond the Google News cycle. Do not watch only the next model announcement or product demonstration. Watch the financial mechanism behind it.

Track operating margin first, separately disclosed AI revenue second, and free cash flow third. Together, those figures will show whether Meta’s advertising machine can finance its ambition without surrendering its economics.