Meta’s second-quarter earnings report showed revenue of $60.8 billion and free cash flow of $784 million. Twelve months earlier, free cash flow was $8.55 billion. The difference, $7.7 billion, went into AI infrastructure across construction sites in Texas, Louisiana, and Alberta.
Mark Zuckerberg used the July 29, 2026 earnings call to defend the spending. He told investors that within five years, billions of people will have a personal AI agent that knows their goals and works on their behalf around the clock. A chief of staff for every person on earth, built on top of apps most of them already use daily.
Whether the numbers being spent right now can lead there remains the open question.
What Zuckerberg Actually Said
According to Meta’s Q2 2026 earnings release, revenue came in at $60.80 billion, up 28% year-over-year, while total costs and expenses climbed 55% to $42.03 billion. On the call, Zuckerberg told investors he believes it’s “extremely unlikely” that billions of people will not have a personal AI agent within five years. He described a future where artificial intelligence works around the clock to manage health, finances, relationships, and household tasks.
Zuckerberg acknowledged that personal AI agents do not yet exist at the scale he envisions, but said widespread adoption is inevitable. He told investors: “I think this is a very large opportunity and one that I think is almost inevitable that someone does. And I think it really plays to Meta’s strengths as a company.”
Zuckerberg has cast personal AI agents as the next mass-market computing platform, and Meta’s apps, WhatsApp and Messenger, as the place most people will meet them. WhatsApp alone has over 200 million small business users, and Meta’s Family of Apps reached 3.60 billion daily active people in June 2026, according to the Q2 2026 earnings release. If personal AI agents become standard, whoever owns the messaging layer owns the relationship.
Zuckerberg said Meta sees selling intelligence as a higher-margin business than selling computing power directly, while also seeing an opportunity in compute. He said personal agents under development at Meta would be “the foundation for our next wave of products and revenue lines in the months and years ahead.”
What the Vision Looks Like in Practice
The agents Zuckerberg described aren’t just smarter chatbots. He said on the call that these agents would understand users’ goals and work on their behalf around the clock, across finances, health, relationships, and household management. Think less “ask it a question and wait for an answer,” more “hand it your to-do list and come back when it’s done.”
WhatsApp is already central to that plan. Zuckerberg said WhatsApp is already the leading platform where users interact with Meta AI, which gives the company a running start on understanding how people actually want to use these tools in daily life rather than just in a tech demo.
The business-facing version of this is already live. More than one million businesses are using Meta’s Business Agent on WhatsApp and Messenger to respond to customers around the clock. There are more than one billion active threads with businesses on WhatsApp, Messenger, and Instagram every day. The agent fields customer inquiries, suggests products from a business catalog, schedules appointments, vets potential sales leads, and completes transactions.
The transition from business agent to personal agent is, in Zuckerberg’s view, a natural extension of the same infrastructure. Once the models are capable enough and the distribution is already in place, the jump from “AI that runs a small business’s customer service” to “AI that manages your personal calendar and finances” becomes a product decision rather than a technical one.
The Financial Picture Behind the Bet
The gap between Meta’s record revenue and its near-vanishing free cash flow tells the story of what it costs to build toward that vision. Meta’s Q2 2026 earnings showed revenue of $60.8 billion beating analyst expectations, but free cash flow collapsed 91% to $784 million as AI infrastructure spending hit $31 billion in one quarter.
Meta guided full-year 2026 capital expenditure to $130 billion to $145 billion, narrowing the floor from a prior range of $125 billion to $145 billion. At the midpoint of $137.5 billion, and assuming Meta’s operating cash flow runs at roughly the Q1 and Q2 pace, minimal room remains for free cash flow over the full year. Meta has issued approximately $25 billion in new long-term debt in Q2 alone to supplement operating cash in funding construction.
Net income for the quarter came in at $15.8 billion and diluted earnings per share of $6.18 fell short of analyst expectations as total expenses climbed 55% to $42.0 billion. Meta’s stock fell almost 10% after the earnings report, while Reality Labs, the division behind its AR glasses and VR headsets, posted a loss of about $4.6 billion for the quarter.
For investors watching the cash flow figure collapse while Zuckerberg describes a vision that won’t pay off for years, the question isn’t whether the vision is coherent. It’s whether the company can afford to keep spending at this pace long enough to find out if he’s right.
The El Paso Bet
Meta and BlackRock announced a venture to develop and operate a data center campus in El Paso, Texas, a project that would cost about $14 billion in development. BlackRock-managed funds will take an 80% ownership stake in the venture, with Meta retaining the remaining 20%, and a portion of BlackRock’s investment will be financed through $12.5 billion in debt.
The El Paso data center campus, already under construction, is designed to provide one gigawatt of compute capacity. Operations are expected to commence in 2028. Meta’s investment in the El Paso project exceeds $10 billion, supporting more than 4,000 construction jobs at peak and 300 operational jobs once complete, with over 2,300 workers already on site.
The structure of the deal is telling. By selling an 80% stake to BlackRock while retaining full operational control, Meta gets access to a gigawatt of compute capacity without having to fund the entire campus directly. BlackRock and its investment platforms gain exposure to long-lived infrastructure, with Meta as the initial tenant.
Meta has said it plans to invest $600 billion in AI infrastructure, including data centers and new jobs in the U.S. by 2028, with an aim to fast-track work on personal superintelligence, which could help spin up new cash flows from the Meta AI app, ad tools, and smart glasses. El Paso is one piece of that. There’s also a $9 billion data center in Alberta, Canada, a gigawatt-scale project in rural Louisiana, and 28 U.S. data centers either in operation or under construction.
What “Working on Your Behalf” Actually Means
Zuckerberg’s language is worth paying attention to. Not “a better assistant.” Not “a smarter search.” An agent that “understands your goals” and works “on your behalf 24/7.” That implies something current AI tools don’t yet do: persistent context, real-world action, and ongoing initiative rather than waiting to be asked.
A personal AI agent in the full sense would need to know that you’ve been trying to reduce spending on subscriptions since March, that your mother’s birthday is coming up and you tend to forget, that your health insurance renewal is next week and you’ve been meaning to compare plans, and that your inbox has three unanswered messages from someone you actually want to hear from. It would handle or flag each of those things without being told to. The difference between that and a chatbot is the difference between an assistant who knows you and a search bar that listens.
That’s a harder engineering problem than building a capable language model, because it requires persistent memory, tool access, reliable judgment about what to act on versus what to surface, and privacy architecture people will actually trust. None of those are solved problems. Zuckerberg didn’t claim they were. He said someone will solve them within five years, and that Meta is better positioned than anyone else to be that someone.
Whether he’s right about the timing is genuinely unknowable. He was wrong about the metaverse timeline, or at least early. He was right about mobile being the next platform when many people weren’t. The pattern is someone who thinks in generational bets and has the company’s finances behind him to let them run.
The Business Case
Zuckerberg’s vision for personal AI agents isn’t purely philanthropic. If Meta’s apps are where billions of people interact with their personal AI agents every day, Meta sits at the center of an enormous amount of information about what people want, what they’re worried about, what they’re planning, and what they’re buying.
That’s a more valuable position than running the world’s largest social network, because the social network shows you what people share. An agent that manages your finances, health calendar, and household knows what people actually do. The difference between those two things is where the next generation of targeted advertising, subscription services, and commerce infrastructure gets built.
Zuckerberg used the earnings call to lay out AI ambitions spanning enterprise services, personal assistants, and faster app development, noting that Meta sees a large enterprise opportunity beyond the business AI agent it launched in June, including APIs, business agents, and potentially selling compute directly to large customers. Selling intelligence at margin while also selling compute capacity to enterprise customers is a version of the business Amazon built with AWS, and it worked. Meta is trying to construct a similar recurring revenue layer underneath the advertising business it already has.
The investor skepticism after the Q2 results wasn’t about whether the vision makes sense. It was about the timeline. Spending $130 to $145 billion in a single year on infrastructure, while free cash flow sits at $784 million for the quarter, requires a level of patience that markets don’t typically reward in the short term. Zuckerberg has bet the company’s financial comfort on the idea that the payoff is coming. Given what he’s already built from much smaller starting points, dismissing that bet entirely would be a mistake. Holding it with some skepticism, on the other hand, seems like the more honest position.
Five Years Is Both Soon and a Long Time
The data centers are going up regardless. The agents are coming, in some form, from Meta or from its competitors. What remains genuinely open is whether the version Zuckerberg described on July 29, billions of people, five years, understanding your goals, working on your behalf, arrives on that timeline, or whether it turns out to be another vision that was right about the destination and wrong about how long the road actually is.
There’s a real difference between a prediction and a plan. Zuckerberg has laid out the prediction with conviction. The plan, so far, is: spend more than any company in history has spent on infrastructure in a single year, build the distribution through apps people already use daily, and trust that the engineering problems will be solved in time. That might be exactly right. The $130 billion being deployed in 2026 alone will eventually either look like the shrewdest capital allocation of the decade, or the most expensive reminder that timelines for transformative technology almost always stretch longer than the people building it expect.
What makes this particular bet different from the metaverse is that the underlying technology is already doing visible, useful things for real people. Personal AI agents as Zuckerberg describes them are further out. But the direction of travel isn’t in dispute. The only honest question left is whether five years from now you’ll be using Meta’s version of this, someone else’s, or something that looks completely different from anything being described today.
Disclaimer: This information is not intended to be a substitute for professional medical advice, diagnosis, or treatment and is for information only. Always seek the advice of your physician or another qualified health provider with any questions about your medical condition and/or current medication. Do not disregard professional medical advice or delay seeking advice or treatment because of something you have read here.
AI Disclaimer: This article was created with the assistance of AI tools and reviewed by a human editor.