Meta’s AI bet cuts free cash flow to $784 million

Meta’s free cash flow fell from $8.55 billion a year ago to just $784 million in the past quarter, as the company’s heavy AI spending began to weigh on its financial metrics. Executives also said full-year revenue is likely to come in below market expectations, without providing a more specific forecast in the earnings report.

Meta’s AI spending hits free cash flow

Meta is the second major AI hyperscaler in a week to face investor concern over cash generation. Google reported negative free cash flow last week, the first such result in the company’s history, after pouring money into AI infrastructure.

Both companies argue that their combined investments—running into the trillions of dollars across the industry—are necessary to meet rising demand. Analysts and other experts, however, are increasingly questioning how quickly that demand will materialize. If spending accelerates faster than revenue, the mismatch could fuel concerns about an AI bubble.

Mark Zuckerberg acknowledged the scale of the commitment during Meta’s earnings call:

Recommended reading

Microsoft’s Anthropic stake gains $3.2B as OpenAI slips

“I get that this is sort of a big bet across the industry. My personal bet is that the people who invest in this are going to be rewarded and feel very good over time.”

Meta says its AI investment is already producing operational benefits. The company is using large language models to improve ad rankings and recommendation algorithms across its social platforms. Susan Li, Meta’s CFO, said every public Reels and Feed post on Instagram is now automatically processed through an LLM and analyzed for characteristics including topic and tone.

“Earlier this year, we reached a milestone of every public Reels and Feed post on Instagram being automatically processed through an LLM and analyzed across dimensions from topic to tone.”

Meta promises agents, glasses and consumer AI

Executives also pointed to a product pipeline that includes Meta glasses and AI agents designed to work continuously on a user’s behalf. Zuckerberg said agents must move beyond their early success in coding and become simple enough for billions of consumers to use without extensive setup.

“The first domain that agents have really taken off in is coding, but engineers are more technical and willing to spend time making those agents work. So, to build great personal agents, this needs to be a great consumer product that just works out of the box and is easy enough for billions of people to adopt and use. I’m very excited about this, and we’re going to have more to share soon.”

The earnings call did not specify when those additional products or announcements would arrive. Zuckerberg said Meta believes its current AI investment is “paying off,” but the company’s cash-flow results show the cost of building that infrastructure is arriving before the financial return is clear.

Meta faces $2.4 billion in legal charges

AI spending was not Meta’s only financial problem. The company recorded $2.4 billion in charges related to legal proceedings during the quarter while fighting lawsuits on several fronts.

One set of allegations claims Meta used discriminatory AI to help determine who would be laid off during a major restructuring earlier this year. The larger group of cases concerns the effects of Meta’s social media platforms on children and teenagers.

Plaintiffs allege that Meta knowingly built addictive design features intended to hook children from a young age, contributing to poorer mental-health outcomes later. California, New Jersey, Colorado and Kentucky are bringing one social-media addiction case against the company.

Earlier this month, Meta disclosed in a court filing that claims brought by the four states could result in $1.4 trillion in damages. As of Wednesday night, Meta’s market valuation was slightly below $1.5 trillion—leaving the potential damages claim close to the company’s entire value.