Meta is spending at a pace that would have looked extreme even by Big Tech standards a few years ago.
The Facebook and Instagram owner has narrowed its 2026 capital expenditure forecast to between $130 billion and $145 billion. Its previous guidance ranged from $125 billion to $145 billion, meaning Meta did not lift the ceiling. It simply decided the lower figure was no longer realistic.
Most of that money will support data centers, servers, networking equipment and other infrastructure needed to train and run increasingly powerful artificial intelligence systems. Meta is not pulling back from the AI race. It is pouring more concrete.
Meta’s AI Infrastructure Spending Is Accelerating
The revised forecast raises the minimum Meta expects to spend this year by $5 billion.
Meta recorded $31.08 billion in capital expenditures during the second quarter of 2026, almost twice the amount reported during the same period a year earlier. The company is expanding its computing capacity as AI becomes more deeply embedded across Facebook, Instagram, WhatsApp and its emerging business products.
That infrastructure is expensive long before it produces a finished product. Meta needs land, power agreements, chips, cooling systems and large data centers capable of supporting AI training and inference at enormous scale.
The spending is not limited to one flagship chatbot or model. Meta wants AI to improve advertising, recommendations, content creation, messaging tools, smart glasses and automated services for businesses.
Free Cash Flow Fell to Just $784 Million
The uncomfortable number in Meta’s results was not revenue. It was free cash flow.
Meta generated only $784 million in free cash flow during the second quarter, down from $8.55 billion a year earlier. That represents a decline of roughly 91%. Heavy infrastructure spending absorbed nearly all the cash left after the company funded its day-to-day operations.
For a company of Meta’s size, $784 million is unusually thin.
It does not mean the company is running out of money. Meta ended June with approximately $90.26 billion in cash, cash equivalents and marketable securities. Still, the collapse in free cash flow shows how quickly the AI buildout is reshaping the company’s finances.
Investors are now being asked to accept a trade: less cash today in exchange for a much larger AI business later.
Revenue Growth Could Not Hide the Rising Costs
Meta’s advertising machine remains strong.
Second-quarter revenue climbed 28% year over year to $60.8 billion, helped by continued growth across its advertising platforms. Meta’s family of apps also reached around 3.6 billion daily active people.
Yet higher revenue did not translate into higher profit.
Net income fell 14% to $15.85 billion as total expenses jumped. Legal costs, employee severance and infrastructure investments all contributed to the pressure. Meta reported roughly $2.4 billion in legal expenses and $1.18 billion in severance costs during the quarter.
The result is a company growing quickly while becoming much more expensive to operate.
Zuckerberg Is Betting AI Will Strengthen the Core Business
Mark Zuckerberg is not presenting AI as a distant research project.
Meta says artificial intelligence already improves the systems that select advertisements, recommend content and keep users scrolling through its apps. Better recommendations can increase engagement. Better advertising models can help businesses target customers more accurately. Both feed directly into Meta’s existing revenue engine.
“AI is accelerating our core business today,” Zuckerberg said as the company released its quarterly results. He also pointed to new consumer products and enterprise opportunities that could emerge from the same technology.
That distinction matters. Meta does not need to create an entirely separate AI subscription business to justify every dollar of spending. Even modest improvements to advertising performance can produce significant revenue when applied across billions of users.
The company is also testing a broader future involving personal AI agents, business assistants and computing services sold to outside organisations. Those ideas could become major businesses. For now, they remain promises attached to a very large infrastructure bill.
Meta Sees Enterprise AI as a New Revenue Opportunity
Meta is also looking beyond its traditional consumer platforms.
The company has discussed offering AI application programming interfaces, business agents and productivity tools to enterprise customers. More than one million businesses are already using Meta-powered AI agents each week across WhatsApp, Messenger and Instagram, according to reports surrounding the earnings call.
This opens another route to monetisation.
A business could use Meta’s AI to answer customer questions, manage conversations, recommend products or automate routine support work. Meta already controls the messaging platforms where many of those interactions happen. It does not need to persuade businesses to adopt an unfamiliar communication channel first.
The harder part will be proving that these tools are reliable enough for serious commercial use.
The Data Center Race Is Becoming Harder to Ignore
Meta is not alone in spending heavily on AI infrastructure.
Microsoft, Alphabet, Amazon and other technology companies are committing huge sums to servers and data centers. The industry believes demand for AI computing will continue rising as models become more capable and companies deploy them across more products.
The risk is that construction may move faster than monetisation.
AI data centers take years to plan and build. Hardware can become outdated quickly. Electricity and cooling costs do not disappear when demand slows. Once companies commit to large projects, reversing course is difficult and expensive.
Meta’s narrowed forecast suggests management sees little room to spend below $130 billion this year. The infrastructure commitments are no longer theoretical. Much of the money is already tied to projects, equipment and contracts.
Investors Want Proof That the Spending Will Pay Off
Meta’s core business remains profitable, but Wall Street is becoming less patient with open-ended AI budgets.
The company now expects total 2026 expenses of between $165 billion and $169 billion. At the same time, its free cash flow has almost vanished for the quarter. That combination makes the return on Meta’s AI spending harder to avoid discussing.
The eventual payoff could arrive through better advertising, paid business agents, personal AI services or access to Meta’s computing infrastructure. Perhaps several of those areas will generate revenue at once.
None is guaranteed to scale as quickly as the spending.
Meta has survived expensive long-term bets before. Its advertising business gives the company room that smaller AI developers do not have. Even so, a capital expenditure range of $130 billion to $145 billion changes the conversation. AI is no longer a side project attached to Meta’s social networks.
It is becoming the machinery underneath the entire company.
What Comes Next for Meta’s AI Buildout
The next few quarters will show whether Meta can keep revenue growing while its infrastructure costs continue climbing.
Investors will be watching advertising performance, new enterprise products and any evidence that AI agents are turning into dependable businesses. Free cash flow will matter too. A temporary decline during a construction surge is one thing. A prolonged squeeze without clear new revenue would be more difficult to defend.
Meta has already made its choice. It wants enough computing capacity to compete at the front of the AI market, even if that means spending most of the cash its operations generate.
The question is no longer whether Meta will build.
It is whether the products arrive quickly enough to fill what it is building.
Sources
- Reuters – Meta Narrows Annual Capex Forecast as AI Buildout Grows
- Meta Investor Relations – Second Quarter 2026 Results
- Associated Press – Meta Q2 Profit Declines as Costs Rise