Meta Stock After Q2 2026: AI Is Working—But the Capital Bill Is Becoming the Story

Data status: August 23, 2026. Meta’s Q2 2026 results answered one question and created another. Artificial intelligence is clearly improving the core advertising machine. Revenue increased 28% to $60.8 billion, ad impressions rose 14%, average price per ad increased 12%, and Family daily active people reached 3.60 billion. The new question for Meta stock is whether those gains can outrun an infrastructure bill that is becoming almost as important as the advertising business itself.

Capital expenditures reached $31.1 billion in the quarter, and full-year 2026 capex guidance stands at $130 billion to $145 billion. Operating cash flow was $31.9 billion, yet reported free cash flow was only $784 million. Meta is not struggling to generate cash. It is deliberately converting an enormous amount of cash into data centers, accelerators, networking, power and long-lived AI infrastructure.

Q2 2026 in numbers

• Revenue: $60.8 billion, up 28%.
• Operating income: $18.8 billion.
• Operating margin: 31%.
• Net income: $15.8 billion.
• Diluted EPS: $6.18.
• Family daily active people: 3.60 billion, up 3%.
• Ad impressions: up 14%.
• Average price per ad: up 12%.
• Capital expenditures: $31.1 billion.
• Free cash flow: $784 million.

The advertising engine is doing almost everything I would want. User growth is modest because the installed audience is already enormous, but monetization per user continues to increase. Both ad volume and price improved at the same time, which is a much stronger signal than growth driven by only one of those variables.

AI is already monetizing inside the existing business

Meta does not need to launch a separate AI subscription for artificial intelligence to create value. Recommendation systems can increase the amount of relevant content a user sees, lengthen sessions, improve ad placement and raise conversion rates. Those gains flow into the existing revenue engine.

This distribution advantage matters. A startup may build a good AI product and still need years to acquire users. Meta can place new recommendation systems, assistants and creation tools in front of billions of people. The customer-acquisition cost is already paid.

For advertisers, the value proposition is even simpler. They do not care whether the system uses a fashionable model architecture. They care whether one dollar of spend produces more revenue. If Meta’s AI improves targeting, creative generation or campaign optimization, monetization can rise without users paying anything directly.

The operating-margin decline needs context

Operating margin fell to 31% from much higher levels a year earlier. Q2 included legal charges and severance expenses, so the reported figure understates the underlying advertising profitability. Even after normalizing unusual items, however, the direction is clear: Meta is spending aggressively.

That is not automatically a negative. A dominant platform should invest when expected returns are high. The analytical problem is that the historical advertising business was unusually asset-light. AI inference and training require physical infrastructure. Every incremental layer of compute makes the company more capital intensive than the Facebook of ten years ago.

Capex is now the central variable

At $130 billion to $145 billion of expected 2026 capital expenditure, small changes in return on invested capital matter enormously. Some spending improves existing recommendation and advertising systems. Some funds frontier-model research. Some supports new products whose revenue is uncertain. Some simply creates enough capacity to avoid being constrained later.

I would not assign the same return assumption to every dollar. Infrastructure tied directly to higher ad conversion has an observable feedback loop. Frontier research has a much wider range of outcomes. The stock deserves a premium only if the combined return on these investments remains comfortably above the company’s cost of capital.

Why free cash flow looks weak despite a strong business

Meta generated $31.9 billion of operating cash flow and spent $31.1 billion on capital expenditures in Q2. That produced only $784 million of reported free cash flow. The number looks alarming until you separate operating weakness from reinvestment.

The operating business is not starved for cash. Management is choosing to reinvest almost all of it. That distinction matters because capex can create future earning power. It also means investors should stop valuing Meta as if historical free-cash-flow margins will automatically return without evidence that the buildout matures.

Depreciation will arrive after the cash spending

Capital expenditures hit cash flow immediately but reach the income statement over time through depreciation. During a rapid buildout, this creates a timing mismatch. Free cash flow can fall first, while depreciation expense continues rising in later periods even if capex growth eventually slows.

AI accelerators also have shorter economic lives than many traditional data-center assets. A building can remain useful for decades. A GPU can become economically outdated much faster when a new generation offers dramatically better performance per watt. The more hardware-intense Meta becomes, the more technological depreciation matters.

Meta AI can be a product, a feature, or both

The clearest returns today come from AI as a feature inside existing products. Better recommendations, ad ranking, content creation and messaging assistance can improve revenue without requiring a new business model.

The larger upside would come from AI becoming a distinct revenue pool. Business agents, creator tools, commerce assistants and new forms of paid messaging could expand monetization beyond advertising. I do not need those products to work for the current business to remain attractive, but they matter for the bull case because they could justify part of the infrastructure build with incremental revenue rather than defensive spending.

WhatsApp remains under-monetized relative to its scale

WhatsApp is strategically important because it gives Meta a global communication layer outside the traditional feed. Business messaging, customer support and transaction-oriented assistants could turn conversations into commercial workflows.

AI can make that monetization more useful. A small business that cannot staff twenty-four-hour support may use an agent inside WhatsApp. A large enterprise may automate routine service while escalating complex cases to humans. Meta already controls the distribution surface where those interactions happen.

Reality Labs remains an expensive option

Reality Labs has represented Meta’s willingness to finance a future interface beyond smartphones. Historically, that meant virtual and augmented reality. In an AI world, glasses and ambient computing may become more strategically connected to the core company because an assistant becomes more useful when it can see, hear and respond without requiring a phone screen.

I still treat Reality Labs as an option rather than part of the core valuation. The losses are real today, while the future economics remain uncertain. If smart glasses become an important AI interface, the investment could look prescient. Until then, the advertising business is paying for the experiment.

Infrastructure financing can change the capital profile

Meta has explored partnerships around large data-center projects. These structures can bring long-duration infrastructure capital into projects without forcing every dollar through the same corporate capex channel. That can preserve flexibility and potentially lower the effective cost of capital.

Investors should still look through the structure. A lease or venture can move accounting presentation without eliminating the economic obligation. I would track total commitments, utilization and the cost of capacity rather than celebrate lower headline capex if financing simply moves elsewhere.

The moat is distribution plus data feedback

Meta’s durable advantage is not one model. It is the combination of billions of users, advertiser demand, engagement data and a closed feedback loop between recommendations and monetization. More interactions create more signals; better ranking can create more engagement; better conversion attracts more advertiser spend.

The risk is that regulation reduces Meta’s ability to use data or that platform gatekeepers such as Apple and Google restrict tracking and distribution. Meta has already shown it can adapt to major privacy changes, but that does not remove the structural dependence on mobile ecosystems it does not fully control.

Valuation framework

The spread between a 22% and 32% normalized free-cash-flow margin is enormous on Meta’s revenue base. That is why the investment debate should focus less on one quarter’s EPS and more on the long-run cash return from the current capital cycle.

What I would watch next

1. Ad impressions and price per ad.
2. Capex growth relative to revenue growth.
3. Free-cash-flow recovery.
4. Depreciation growth.
5. Business messaging monetization.
6. AI product revenue outside ad optimization.
7. Reality Labs operating losses.
8. Net debt and infrastructure commitments.

What would make me more bullish?

I would become more constructive if advertising continues to grow at a strong double-digit rate while capex growth slows, free cash flow rebuilds and Meta demonstrates direct monetization from AI products or business messaging. Evidence that infrastructure partnerships lower the cost of expansion without sacrificing control would also improve the thesis.

What would break the thesis?

The thesis weakens if ad pricing decelerates while capital spending remains close to current levels, if inference costs rise faster than monetization, or if regulatory changes materially limit Meta’s ability to use data across its platforms. The worst outcome would be a successful AI engagement strategy that benefits users but produces mediocre returns on the capital required to support it.

My conclusion

Meta’s Q2 2026 results prove that AI is already working inside the core business. Revenue rose 28%, ad impressions increased and pricing improved. This is not a speculative AI story.

The uncertainty sits in the capital structure of that success. Meta is spending at a scale that changes the nature of the company from an exceptionally asset-light platform into a digital business with a large physical infrastructure layer.

I remain positive on the business, but I want evidence that the capital cycle eventually produces stronger free cash flow rather than simply a larger asset base. AI has already improved the product. The next test is whether it improves owner economics after paying for the machines that make it possible.

Primary sources

• Meta Q2 2026 results
• Meta investor financials

This article is analysis, not investment advice.

The real question is return on AI capital

Meta’s capital program is now so large that investors should think about it with an industrial mindset. The company is converting cash generated by advertising into physical infrastructure. That means future returns depend on utilization, depreciation, energy costs and the pace at which AI improvements translate into revenue.

The best case is powerful: better recommendations increase engagement, better ad systems raise conversion, and the same infrastructure supports new products. The weaker case is that Meta spends enormous amounts merely to preserve competitive parity.

What would prove the capex is earning its keep?

I would look for three signals. First, advertising growth should remain stronger than the broader digital-ad market. Second, free-cash-flow margins should recover as the infrastructure build matures. Third, AI should create new monetization beyond optimization of existing ads, especially in business messaging and assistants.

If all three occur, the capital cycle can be viewed as offensive investment. If only the first occurs, the spending may be partly defensive.

Why WhatsApp could matter more than investors expect

WhatsApp combines global distribution, business communication and a natural environment for conversational agents. A merchant can use an AI agent for support, product discovery, appointment scheduling or order updates inside the same thread where the customer already communicates.

This creates a path to monetization that does not depend on adding more ads to social feeds. It can also deepen switching costs for businesses if messaging, support and commerce workflows become integrated.

AI agents introduce a new security problem

As assistants gain the ability to take actions, permissioning becomes more important. A model that can draft text is one thing. A model that can send money, change an ad campaign or access customer data creates operational risk. Meta will need strong controls, audit trails and trust if business agents become a meaningful product category.

Why open models can still support a closed economic moat

Meta’s model strategy does not need to monetize every model directly. Open distribution can strengthen the developer ecosystem, influence standards and reduce dependence on competitors. The economic moat can remain closed at the distribution layer: Instagram, WhatsApp, Facebook and advertising demand.

This is a useful distinction. Open technology can still create value for a company that controls the customer interface.

How I would normalize free cash flow

Current free cash flow is depressed by extraordinary investment. I would not simply capitalize today’s quarterly number. I would estimate a mature capex level after the current buildout, then subtract a realistic depreciation and maintenance requirement. That produces a more useful normalized owner-earnings figure.

The key is not to assume capex returns to old pre-AI levels. Meta has become structurally more capital intensive. A normalized margin should reflect that.

Related reading on The Kapital

Investors comparing AI capital cycles should also read our Amazon AWS and AI capex analysis and our guide to how higher bond yields affect growth-stock valuations.

FAQ

Is Meta’s AI spending already producing revenue?

Yes, indirectly. Better recommendation and ad-ranking systems appear to be improving engagement and monetization inside the existing advertising business.

Why is free cash flow so weak despite strong earnings?

Because Meta is spending an unusually large amount on data centers, accelerators, networking and related infrastructure.

Is Reality Labs still central to the thesis?

I treat it as optionality rather than core value. The advertising and messaging businesses remain the foundation of the investment case.

What would be the biggest warning sign?

Advertising growth slowing while capex remains elevated and free cash flow fails to recover.

Does a high capex number automatically make the stock unattractive?

No. High capex is valuable if the returns on that capital are strong. The question is return on invested capital, not spending by itself.

Meta’s AI economics have two different time horizons

The first horizon is already visible: recommendation and ad-ranking improvements. These can monetize immediately because they improve an existing business with billions of users and advertisers. The second horizon is much more uncertain: assistants, agents, smart glasses and new computing interfaces.

Investors should not value both horizons with the same confidence. The core advertising benefits deserve a higher probability. New product categories should be treated as optionality until revenue and retention become observable.

Why depreciation can surprise investors later

During a rapid buildout, capex dominates the cash-flow discussion. Several years later, depreciation can become the more important accounting pressure. If Meta installs huge amounts of hardware today, the income statement will absorb that cost over future periods.

This matters because a slowdown in capex growth does not immediately restore operating margins. The installed asset base continues to depreciate.

Energy and power availability are becoming strategic constraints

AI infrastructure needs enormous amounts of electricity. Data-center expansion is therefore constrained not only by chips but also by grid connections, generation capacity and local permitting. Meta’s ability to secure long-term power can become part of the competitive advantage.

The same issue raises cost risk. If power prices rise or projects are delayed, expected returns on infrastructure can fall.

Why advertising remains the valuation anchor

Despite the AI narrative, Meta is still fundamentally an advertising company. The core valuation should therefore begin with engagement, ad load, conversion, pricing and advertiser return on spend. AI matters because it improves those variables.

If advertising economics remain strong, the company can finance ambitious future bets. If the ad engine weakens, the same bets become much more expensive.

What a successful capital cycle looks like

A successful capital cycle would show slowing capex growth, stable or improving utilization, persistent double-digit advertising growth and a recovery in free-cash-flow margin. The company would emerge with a larger infrastructure base and stronger monetization without permanently lowering owner returns.

What an unsuccessful cycle looks like

An unsuccessful cycle would show continued heavy investment while incremental revenue growth slows and depreciation rises. In that case, AI could improve the product while reducing the economics of the stock.

Final investor checklist

• Is ad pricing still rising?
• Are impressions growing without damaging user experience?
• Is capex growth slowing relative to revenue?
• Is free cash flow recovering?
• Are AI products creating revenue beyond ad optimization?
• Are Reality Labs losses controlled relative to core cash generation?