TL;DR
- Revenue Forecast: Meta forecast third-quarter revenue of $61 billion to $64 billion.
- Share Movement: Meta shares were reported down as much as 10 percent after hours on July 29.
- Spending Outlook: Meta’s 2026 capital-expenditure range now runs from $130 billion to $145 billion.
- Historical Cash: Meta’s second-quarter 2026 revenue reached $60.801 billion, but free cash flow was $784 million.
- Payoff Routes: Mark Zuckerberg, Meta’s founder and CEO, points to AI agents and computing leases, but neither has shown material revenue.
Meta has issued a revenue forecast of $61 billion to $64 billion for the third quarter, while its shares fell as much as 10 percent after hours. Meta founder and CEO Mark Zuckerberg defended the rising artificial intelligence spending as investors weighed that outlook against the cost of data centers and chips.
Zuckerberg’s defense rests on AI helping the core business now and new products following. Product adoption and outside demand for the company’s computing infrastructure have not yet demonstrated a material return. Meta’s AI case remains dependent on future value rather than measurable AI income.
The Spending Test Behind the Revenue Beat
Meta’s July 2026 results raised the lower end of its capital-expenditure outlook through a $125 billion-$130 billion shift and $145 billion ceiling. Capital expenditure is long-term spending on infrastructure such as data centers and chips. Earlier AI infrastructure spending had drawn investor scrutiny. But Meta’s separate total-expense outlook also includes operating and other costs beyond the infrastructure budget.
Meta’s completed quarter sharpened that tension with $60.801 billion revenue and only $784 million cash flow. The two metrics answer different questions: revenue measures what Meta’s businesses brought in, while free cash flow captures what remained after operations and investment.
Advertising still supplied about 98% of second-quarter revenue. Meta’s established advertising operation is financing chips, servers, and data centers while newer AI channels develop. Infrastructure must be paid for before those services can replenish the cash committed to them, making the timing of returns the central investor concern.
Money committed to long-lived infrastructure is unavailable for buybacks, acquisitions, or other corporate uses, even when advertising continues to grow. Strong advertising sales can coexist with a tighter cash position when the company commits heavily to AI infrastructure. Meta is asking investors to accept that near-term constraint before newer AI businesses produce measurable income.
Minda Smiley, a senior analyst at Emarketer, offered a blunt assessment: “Meta’s strong revenue growth will once again be overshadowed by its capital expenditure projections.” Her assessment explains why the revenue beat did not settle the debate. Higher investment can be defensible, but Meta still needs to connect infrastructure outlays to durable products or measurable income.
AI could support Meta’s finances in two ways: improving the advertising operation or generating direct income from newer services. Current figures do not separate those contributions, preventing investors from determining whether AI improves ad economics, generates standalone sales, or mainly increases costs. Advertising’s dominant share leaves agents and computing leases facing a higher burden of proof as standalone businesses.
Where Meta Says the Payoff Could Come From
CEO Mark Zuckerberg said companies had made offers for Meta’s computing power at a meaningful premium over its acquisition cost and described Meta’s lease-versus-keep calculation. Computing capacity means the chips and data-center systems used to train and run AI. Leasing some of it could generate income, while retaining it would support Meta’s own models and services because the same hardware cannot serve both workloads at once.
Meta said that by the end of the second quarter, more than one million businesses had adopted its business agents on WhatsApp and Messenger. The agents are AI assistants that can perform tasks for people or companies. Meta’s AI image-editing tools one the other side now reach nine million small businesses. Meta has not tied either adoption figure to customer payments, service costs, or margins, so user counts cannot establish whether the products cover their service and infrastructure costs. Given the low free cash-flow, most probably not by a large extend.
Meta has started to explore monetizing spare computing capacity as the infrastructure program tightens its cash generation. But each chip committed to an outside customer is then also unavailable for Meta’s own model training and services. Neither adoption counts nor Zuckerberg’s premium-offer claim establishes material revenue or completed contracts, making the next earnings release the test for measurable AI income and sufficient retained capacity.