Meta generated $64.6 billion in revenue during the third quarter of 2026, a 26% increase from a year earlier. Its advertising business was thriving, users were spending more time across its platforms, and the company remained profitable. Wall Street responded by selling the stock.
Shares fell sharply on October 22 as investors absorbed another enormous increase in infrastructure spending. Meta now expects capital expenditures of $135 billion to $145 billion for 2026, while its quarterly free cash flow has slipped below zero. The company that turned Facebook and Instagram into extraordinary cash-generating businesses is now spending heavily enough to consume virtually all the cash its operations produce in a quarter.
Mark Zuckerberg is betting that this infrastructure will secure Meta’s position in artificial intelligence, improve its existing products and create new businesses. There is evidence that parts of the strategy are already working. The problem is that Meta AI spending is growing so quickly that investors can no longer judge the company by advertising revenue alone.
Meta’s $145 billion AI spending plan is bigger than it sounds
The $145 billion figure is the upper end of Meta’s latest 2026 capital expenditure guidance, not an amount the company has already spent. The official range, announced alongside its October 21 earnings, is $135 billion to $145 billion, including principal payments on finance leases. Meta raised the lower end of its previous $130 billion to $145 billion forecast.
Nor is every dollar earmarked exclusively for artificial intelligence. Meta needs infrastructure to operate Facebook, Instagram, WhatsApp and its advertising systems, even without its ambitions in frontier AI. The new investment program includes data centers, servers, networking equipment and computing capacity serving both established products and emerging AI initiatives.
Still, AI is a major reason for the expansion. Training sophisticated models requires enormous computing clusters, while deploying those models to billions of potential users creates continuing demand for processing power. Meta is investing in the infrastructure needed to compete with Google, OpenAI and other AI developers, rather than depending entirely on outside providers.
The scale is visible in the company’s balance sheet. Its September 30 SEC filing lists $133.6 billion in servers and network assets before accumulated depreciation, alongside $89.5 billion in construction in progress. Much of that unfinished construction relates to data centers and technical infrastructure. These are physical investments that take time to build and cannot necessarily be repurposed or sold at their original cost.
The distinction between spending and returns is important. A new server can improve advertising recommendations immediately, support a model-training project that takes months, or provide capacity for a product whose commercial future remains uncertain. All three may be strategically useful, but they have very different financial payback periods.
How Meta can report billions in profit and still run out of free cash flow
The most revealing number in Meta’s third-quarter results may be its negative $2.12 billion free cash flow.
That figure looks strange beside $9.86 billion in reported net income and $32.79 billion in operating cash flow. But the three measures describe different parts of the business.
Net income is an accounting measure. Operating cash flow tracks cash generated through business operations, with adjustments for noncash expenses and working-capital movements. Free cash flow goes further by deducting investment in property and equipment and, under Meta’s definition, principal payments on finance leases.
In the third quarter, Meta generated approximately $32.79 billion from operations. It then spent $33.94 billion purchasing property and equipment and another $964 million on finance-lease principal payments. The result was a $2.12 billion free cash flow deficit.
Meta’s advertising machine is still generating enormous amounts of cash. Its infrastructure program is consuming that cash even faster.
This does not mean Meta is in financial distress. At September 30, the company held $82.45 billion in cash, cash equivalents and marketable securities. It also remained free-cash-flow positive for the first nine months of 2026, generating approximately $11.05 billion, although that was considerably below the $29.51 billion recorded during the comparable period of 2025.
The deterioration matters because it changes how much financial flexibility the business creates internally. Money committed to new infrastructure cannot simultaneously fund acquisitions, shareholder distributions or other investments. If the spending program continues expanding, Meta may increasingly rely on its existing liquidity, borrowing capacity or capital markets to finance it.
The $9.89 billion legal charge changes the earnings picture
There is another complication in the third-quarter numbers: a $9.89 billion charge related to Meta’s agreement with US state attorneys general over litigation involving teen mental health and social media.
The charge pushed reported operating expenses sharply higher and reduced quarterly operating income to $12.99 billion, down from $20.54 billion a year earlier. On the surface, that looks like a dramatic deterioration in profitability.
But the legal expense is not an AI infrastructure cost. Meta says that excluding the charge, operating income would have been $22.88 billion, while diluted earnings per share would have been $7.40 instead of the reported $3.83.
That adjustment is useful for understanding the performance of the underlying business, although the legal obligation remains a real financial issue. Removing an exceptional expense from an analytical comparison does not make the associated liability disappear.
It also highlights why the negative free cash flow figure deserves separate attention. The legal charge affects accounting earnings, while the massive investment program is visible directly in cash expenditure. Conflating the two would exaggerate the deterioration of Meta’s operating business and obscure the actual source of investor anxiety.
Without the litigation charge, Meta’s quarterly operating performance looks substantially stronger. Its capital intensity, however, remains impossible to ignore.
Zuckerberg has a credible argument: AI is already making Meta money
The bearish interpretation of Meta’s results would be incomplete without acknowledging that AI is already contributing to the company’s established business.
Advertising revenue reached $63.06 billion in the third quarter, also up 26% year over year. Across Meta’s Family of Apps, advertising impressions increased 10%, while the average price per ad rose 14%. Daily active people reached approximately 3.64 billion in September, up 3%.
Those figures suggest Meta is extracting more advertising value from its enormous audience, not simply relying on rapid user growth. Better content recommendations can keep people engaged, while improved advertising models can help businesses reach potential customers more efficiently.
Meta explicitly credits AI investments with generating additional advertising revenue through stronger engagement, recommendations and monetization. That claim is supported by the broader improvement in advertising performance, although the company does not provide a precise public breakdown showing how many dollars of incremental revenue came from each AI initiative.
This is a substantial advantage over AI developers that must establish new markets for their technology. Meta can deploy machine learning directly into a business that already serves billions of people and generates tens of billions of dollars in quarterly sales.
Its newer AI ambitions are less financially transparent. Meta is developing frontier models and consumer-facing AI experiences, including the Muse family announced in its earnings release. These projects could eventually produce new revenue streams, improve existing services or reduce internal costs. Their long-term economics remain much harder to measure than improvements in advertising.
The strongest defense of Zuckerberg’s strategy is therefore not that every new AI product will become a standalone blockbuster. It is that AI can make Meta’s existing business more valuable while creating options for future growth.
Why Wall Street is losing patience with the spending
Investors are not rejecting AI investment itself. They are questioning the relationship between the speed of spending and the visibility of returns.
On October 22, Reuters reported that Meta shares were down 6.1% during the session, while the cost of insuring the company’s debt against default reached a record high. Five-year credit default swaps widened to 114.79 basis points, according to Markit data cited by Reuters.
Credit default swaps are financial contracts used to insure against a borrower’s default. Their pricing reflects market perceptions of credit risk, among other factors. A rise does not mean a company is about to default, but it can indicate growing concern about its financial commitments.
Meta’s problem is partly one of timing. Building data centers and purchasing processors requires cash today. The resulting assets may generate economic benefits over several years, while some of the technology can become obsolete much sooner.
Depreciation creates another delayed effect. Capital expenditure does not generally appear as an immediate expense in the income statement. Instead, the cost of an asset is recognized over its estimated useful life. As Meta’s infrastructure base expands, depreciation charges can place increasing pressure on future reported margins.
That pressure is already visible. Meta’s SEC filing shows property-and-equipment depreciation of $6.81 billion in the third quarter, compared with $4.76 billion a year earlier. Continued investment will add to the asset base, although future depreciation will depend on the timing, classification and useful lives of those assets.
There is also an uncomfortable technological risk. AI hardware and model architectures are evolving rapidly. A data center designed around today’s computing requirements might remain useful for years, but some of its equipment could lose economic value faster than originally anticipated.
Meta has the resources to absorb mistakes. What investors want to know is how expensive those mistakes could become.
The new Class C shares add a governance problem
Meta’s earnings announcement contained another development that attracted market attention: the creation of a new publicly listed, non-voting Class C common stock, planned to take effect in January 2027.
The company intends to distribute one Class C share for each existing Class A or Class B share, subject to the stated conditions. The distribution itself is not designed to change shareholders’ voting power. Meta says the new class will provide flexibility for capital raising, employee compensation, acquisitions and other corporate purposes.
The governance implications extend beyond the initial distribution. Future issuance of non-voting shares could allow Meta to raise equity capital without proportionately expanding voting rights. It could also dilute existing shareholders’ economic interests, depending on how such shares are issued.
That is particularly relevant because Zuckerberg already exercises substantial influence over Meta through its voting structure. The company presents continued founder leadership as a strategic advantage, while critics may see the new share class as another mechanism limiting shareholders’ ability to influence major capital-allocation decisions.
Reuters reported that some investors interpreted the move as creating a potential route for additional equity issuance. No such future dilution should be treated as inevitable, but the concern is understandable when the company is simultaneously committing unprecedented resources to infrastructure.
What investors should watch in 2027
The next year will offer a more useful test of Meta’s AI strategy than any single earnings reaction. The company has not provided a full-year 2027 revenue forecast, but it expects continued strong growth and aggressive investment in its core business and emerging AI initiatives.
The first indicator will be capital expenditure guidance. If spending continues rising, investors will want evidence that new infrastructure is supporting revenue growth or improving operating efficiency. A slower spending trajectory could relieve cash-flow pressure, but it might also reflect changing demand or strategic priorities.
Free cash flow will be equally important. Meta’s ability to convert operating cash into money remaining after infrastructure investment will show whether the expansion is becoming more financially sustainable. Operating margins and depreciation trends will help reveal how the enlarged asset base affects reported profitability.
Advertising performance remains the clearest near-term measure of AI’s commercial contribution. Growth in impressions, pricing, engagement and revenue will provide evidence of whether Meta’s machine-learning investments continue improving its most profitable business.
For newer AI products, the crucial missing information is more specific: identifiable revenue, usage economics and the costs of operating those services. Without such disclosures, investors may struggle to separate profitable applications from promising experiments.
Meta also expects Reality Labs losses to decline, forecasting approximately $18 billion to $19 billion for 2026 and around $15 billion in 2027. Progress there could release some resources, but it would not by itself establish the profitability of the company’s broader AI expansion.
Meta’s greatest strength is becoming its biggest financial test
Meta is not a struggling technology company desperately searching for a new business. Its advertising operations remain extraordinarily profitable, its audience is enormous, and its existing AI systems are already contributing to commercial performance.
That financial strength gives Zuckerberg something most AI competitors lack: the ability to fund a vast infrastructure program from an established business rather than depending entirely on outside investors.
But financial strength does not eliminate the need for discipline. Every new data center creates future operating costs, depreciation and pressure to generate sufficient returns. The larger the investment program becomes, the harder it is to justify on strategic ambition alone.
Meta built one of the world’s most profitable advertising machines. The question now is whether that machine can finance the next generation of artificial intelligence without sacrificing the financial discipline that made it so valuable.