Meta Q2 earnings recap: Stock tumbles nearly 10% after profits miss Wall Street estimates

Thanks for joining us. See you next quarter!

Open source AI is an "important part of the ecosystem" and "good for the world," Meta's CEO says. Meta has "always" said that it would mix open and closed models, Zuckerberg adds. "That continues to be true," he says.

Zuckerberg says he wanted to ensure Meta Superintelligence Labs was "uninhibited" in its development process. "We expect that we will get back to releasing some open source models at some point soon," he says.

Zuckerberg says he is "quite happy" with Meta Superintelligence Labs' trajectory. The lab, headed by Scale cofounder Alexandr Wang, released its first model, Muse Spark, in April, followed this month by the image-generation model Muse Image and Muse Spark 1.1, which is designed for coding, computer use, and other agentic tasks.

Meta is now scaling up to larger, more advanced models. Zuckerberg says its vast user data could provide a durable advantage along with a research culture that compounds "consistently with low drama" over time.

There's "nowhere near enough compute for all the demand," Zuckerberg says. It would be "foolish" to sell all the compute and take a short-term profit, he says. "It's always a portfolio."

Meta wants to use its capital to build enough long-term compute, knowing it can monetize it both directly and through its AI, he says.

Zuckerberg says Meta is making a big bet on launching AI agents for everyone, calling it an "extremely important and massive market."

The CEO predicted that in a few years, billions of people will have a personal AI agent that understands their personal and career goals and works for them 24/7.

Let's face it: there's a whole population that only uses AI for search. Bernstein analyst Mark Shmulik asked about this.

"Some things have already broken through," Zuckerberg responds. Coding is first, he says, because it's an "inherently digital and closed-loop activity," he says. Personal AI agents are a "massive market opportunity," Zuckerberg says, and "building for consumers is a little bit different than building for developers."

The slump has doubled since the earnings results crossed.

Meta is developing internal productivity and coding tools, Zuckerberg says, "partially because we need to build them ourselves." There's a "large opportunity," he says.

Zuckerberg says that a "substantial amount of the compute" goes to training Meta's models. The rest goes to other opportunities, he says, from "optimizing and improving" the core business, developer tools, and business agents. Meta can also "sell compute directly" and has "quite a number of offers," he adds.

Selling compute is a "big opportunity," and there is a "significantly higher margin on selling intelligence," Zuckerberg says.

CFO Li warns of "scrutiny on youth-related issues" and coming "youth-related trials." Meta has also referenced these trials in prior earnings calls. The company disclosed this quarter that it had $2.40 billion of charges "related to legal proceedings."

CFO Susan Li said the industry is still "tight" on compute as AI usage soars. She said Meta is geared toward "maximizing 2026 and 2027 capacity" for its AI technology, and that Meta has "high confidence" in its ability to build new foundational models.

Li says that Meta's API is available at a "competitive price," and that the company is "encouraged by initial results." She points to Muse Spark's availability on OpenRouter, an AI routing platform.

Meta is using large language models to "predict the best ad for each person," CFO Susan Li says. Ad-matching is more intelligent and precise, she says, leading to noticeable improvement. She cites some stats: an 8.3% increase in ad clicks, and an 15.7% uplift in conversions on Facebook.

Advertisers have had their critiques of Meta's AI ads push.

The extended stock slump comes as CFO Susan Li nears the end of her prepared remarks. The live Q&A with Wall Street analysts is next up.

Meta's costs jumped 55% from a year ago, and CFO Susan Li says fees for using third-party AI models contributed to the increase. Those fees are based on tokens — the units AI systems use to process information — and they can add up fast.

Meta is hardly alone. Companies across tech are reining in "tokenmaxxing" after finding that bigger AI bills don't always produce more useful work.

AI agents evaluate "content quality," find trends, and test feed rerankings, CFO Suan Li says. Every single public Reels and Feed post on Instagram is automatically processed through an LLM, she adds.

CFO Susan Li says that Reality Labs revenue was up 16% year-over-year. That's driven by Meta's buzzy AI glasses, which have seen "strong growth," Li says. Its Quest virtual reality headset sales were lower, she says.

The Meta CFO is diving into more detail on the quarterly results.

Zuckerberg is positioning Meta as bringing "superintelligent" AI to the masses — and arguing that's a good thing, in contrast to some other executives who have warned about the technology's dangers.

"Before I wrap, I want to mention that I just published an op-ed about why I'm so optimistic that we are building a positive future for everyone," Zuckerberg says. "At Meta, we have always built technology to put power in people's hands, so that they can connect with the people they care about and shape the world in the ways that they want."

Meta is investing "aggressively" in infrastructure, Zuckerberg says. He points to Meta's new venture with BlackRock, a 1 gigawatt data center in El Paso, Texas. Meta is receiving offers for its compute for a "significant premium over what we paid for it," he says.

Meta's big AI reorganization from last year is going well, Zuckerberg says, citing the latest models, Muse Spark 1.1 and Muse Image. Both models were launched last month and are driving a daily 60% increase in interactions with Meta's AI chatbot.

Meta is building personal AI agents that Zuckerberg said will work "24/7 on your behalf" to help with everything from health and relationships to finance and building businesses.

"AI is helping our teams speed up product development," Zuckerberg says. He gave the examples of Instagram Instants and Seller, the Facebook Marketplace app. It will be "a lot easier to ship new apps," he says.

Meta's new Muse Image and Muse Video will "dramatically expand the universe of content" on Meta's platforms, Zuckerberg says. He says that there is a "new and nearly infinite universe of personalized content."

Threads is the "fastest growing conversation app ever," crossing 500 million monthly active users. He also shouts out Instagram, which reached 2 billion daily active users, and WhatsApp, which hit a messaging record.

The behind the scenes of the operator's earnings call setup can be heard.

Mark Zuckerberg and other Meta execs, including CFO Susan Li, are on the call. They'll start by reading through prepared remarks and then will take questions live from Wall Street analysts.

Meta's call kicks off in a minute. The stock is down more than 6% after hours.

The number of people active across Meta's platforms — mainly Facebook, Instagram, and WhatsApp — hit a new all-time high of 3.6 billion in the second quarter. Last quarter, investors worried after Meta disclosed that its user base was starting to decline for the first time. Meta, however, said this was mostly due to government internet blackouts in Iran.

Meta's advertising revenue jumped 27%, but its costs rose 55% as it poured money into AI infrastructure. Free cash flow plunged 91% from a year ago to just $784 million, while capital spending surged 83% to $31.08 billion.

Meta booked $2.4 billion in charges related to legal proceedings and $1.18 billion in severance expenses from its May layoffs. The charges help explain why costs jumped 55% and operating income fell 8%, even as revenue grew 28%.

Meta reported 75,472 employees at the end of June, down just 1% from a year earlier. But that figure still includes about 8,000 workers affected by its May layoffs, most of whom will disappear from the official count by the end of this quarter.

Reality Labs — the division that oversees Meta's hardware efforts — made $431 million in quarterly revenue, up 16% year over year. While that's solid growth, the division remains tiny, generating less than 1% of the cash generated by Meta's ad business.

Advertising revenue rose 27% to $59.36 billion, beating analysts' $59.07 billion estimate. Ad impressions grew 14%, while the average price per ad climbed 12%.

Wall Street expects Meta to deliver $63.17 billion in revenue for the third qurater. That's on the high end of the guidance CFO Susan Li gives: $61-64 billion.

As AI spending balloons, investors and analysts alike are keeping an eye on Meta's future spending plans. CFO Susan Li says Meta expects capital expenditures to be between $130-145 billion for the year, narrowed slightly from its prior outlook. Wall Street expected about $135.79 billion, keeping it squarely within the guidance.

The company reports $42.03 billion in costs and expenses for the quarter — including some additional charges that upped the bill. Meta paid $2.4 billion in charges related to its "legal proceedings" and $1.18 billion in severance expenses after it laid off roughly 10% of its workforce in May.

• Revenue: $60.8 billion vs. $60.23 billion estimate
• EPS: $6.18 vs. $7.14 estimate
• Advertising revenue: $59.36 billion vs. $59.07 billion estimate
• Reality Labs revenue: $431 million vs. $428.7 million estimate
• Capex: $31.08 billion

A key question on investors' minds is whether Meta's AI spending is boosting its all-important digital advertising business. However, Meta may be facing a "trust toll" thanks to its AI-generated advertising tools, said Forrester analyst Mike Proulx.

Business Insider reported earlier this month that its AI ad tools have been causing chaos for some brands. "That's a big problem for a company that depends on ad revenue," Proulx said.

The tech industry's rush for compute isn't slowing down, and Bloomberg recently reported that Meta is developing plans to build its own cloud infrastructure that it could lease to customers.

As Google, Amazon, and Microsoft can attest, it's a good business to be in right now, and one that would help Meta justify its deep investments elsewhere. It's possible we'll hear more about these plans on Meta's earnings call.

Investors love a clean strategy — but right now, Meta is spinning a lot of plates. "Meta spent the quarter cutting jobs, reorganizing around AI, launching new products, testing new apps, expanding subscriptions, pushing further into smart glasses, and pouring billions more into AI infrastructure," said Forrester analyst Mike Proulx.

"Investors should spend less time counting Meta's AI announcements and more time evaluating whether the company can handle this much disruption at once," he added.

Needham is less bullish, maintaining a Hold rating and a more muted $606 price target, about 2% higher than current levels. Analyst Laura Martin says that her team favors waiting to purchase Meta stock until the current capex cycle is over, citing concerns about its plans for growth.

"We retain our HOLD, because we believe that META's strategy diffusion destroys economic value," she wrote. "We think META is asking for too much capital for too many projects, and since competition is intense, strategy dilution lowers the odds that it will succeed at any of them." It cited projects like its Ray-Ban glasses, the Metaverse, and Reality Labs as having uncertain futures in terms of profitability.

Goldman expects Meta to continue ramping up AI capex, but it's not worried. According to analyst Eric Sheridan, Meta's spending will reach $534 billion from 2026 to 2028 as it continues to monetize its AI compute buildout. The bank has a Buy rating on the stock and a bullish $815 price target, a 37% jump from current levels.

"We continue to see sustained operating momentum in terms of our advertising checks across Q2, which bodes well for its forward growth trajectory," he wrote. "We continue to see the application of additional compute to the company's core Family of Apps products as generally underappreciated by investors."