Meta Bets $145B on AI as Debt and Risk Climb

Meta lifted the bottom of its 2026 capital-spending forecast on Wednesday, Oct. 21, raising it to $135 billion from $130 billion. The company now expects to spend between $135 billion and $145 billion this year. That money will go toward data centers, advanced chips and computing power as Meta races OpenAI and Anthropic, with no sign the competition will slow.

The scale of the bet is hard to overstate. Meta's third-quarter capital expenditures reached $34.91 billion, up from $31.08 billion in the prior quarter. Analysts had expected annual spending of $139.79 billion, meaning the company is spending more than Wall Street anticipated. On the "Cheeky Pint" podcast earlier this month, CEO Mark Zuckerberg said he'd been willing to bet up to $200 billion on AI before he knew whether the models would work.

Who Pays the Price

The spending is already squeezing Meta's cash. Free cash flow came in at negative $2.12 billion in the third quarter, compared with positive $10.63 billion a year earlier. The New York Times put the figure at negative $2 billion for the three months ended Sept. 30, down from $784 million at the end of June and $8.55 billion in June 2025. Long-term debt totaled $80.9 billion last quarter, up from $58.7 billion a year ago.

Meta has also used financing that doesn't appear on its balance sheet. Private-credit firms financed a $50 billion Louisiana data-center deal, a structure that keeps obligations out of sight for ordinary investors and regulators tracking the company's exposure. Market commentary described investors as anxious about debt issued by major AI companies and signs of strain in funding structures at CoreWeave and other riskier corners of the boom. The people bearing the downside aren't always the ones making the decision.

The Control Question

Meta said it would issue a new class of non-voting Class C common stock in January 2027, effectively creating a two-for-one split. The company said the shares would give it more ways to raise money, fund acquisitions and retain talent while keeping Zuckerberg's control intact. The New York Times also reported the stock could be used for employee compensation. Shareholders who don't carry voting power now have less say over how that capital gets spent. The split doesn't change who's at the wheel.

Strong Revenue, Heavy Costs

The top line looks healthy. Third-quarter revenue hit $64.60 billion, up 26% from a year earlier and ahead of the $63.37 billion Wall Street expected, according to LSEG data. Fourth-quarter guidance of $73 billion to $76 billion also topped estimates. Costs and expenses, however, rose 68% to $51.61 billion. Profit was $9.8 billion, against $2.7 billion a year earlier, when a tax charge distorted results.

Diluted earnings per share came in at $3.83. Without a $9.89 billion charge tied to a child-safety settlement, that figure would have been $7.40. The Verge reported that income from operations would have been $22.89 billion without the charge.

The settlement traces back to August, when Meta agreed to pay to resolve claims from nearly all U.S. states that its platforms were designed to addict children. The Verge reported the total at $17.1 billion, with $5.3 billion contingent on YouTube and TikTok adopting stricter settings and safeguards for kids. Reuters described the deal as up to $18 billion over a decade, including limits on teenagers' use of Facebook and Instagram. The agreement spared Meta a lengthy trial that could have brought more internal documents about its treatment of young users into public view.

On Wednesday, a New Mexico judge ordered Meta to pay $25 billion after a jury found the company misled state residents about sharing user data with third parties and enforcing rules on hate speech and misinformation. The case related to British political consulting firm Cambridge Analytica. Emarketer senior analyst Minda Smiley called the penalty a "stark reminder of the current legal challenges Meta faces, which will continue into next year." She added: "Questions and concerns related to privacy and safety will only ramp up should AI agents become more mainstream."

Muse and the Consumer Test

Meta's consumer AI agent, Muse, has drawn more than 2 million daily users and 3 million downloads, according to Business Insider. The New York Times reported it became the fastest-growing AI app since OpenAI's ChatGPT launched in 2022. Some users told Business Insider they relied on it to plan trips and prepare for exams. At Meta's Connect keynote in September, Zuckerberg said Muse "has a novel business model." He told the audience: "We believe that Muse will make you money, and we're standing behind this by making Muse free for a huge number of tokens with the expectation that over time, we will profit by taking a small fee from transactions."

Investors cheered the results, with shares rising more than 4% in extended trading, according to Reuters. Business Insider reported a 4.12% after-hours gain. Some enterprises have restricted Muse's access over security concerns. Trust matters for adoption.

Official Response

Zuckerberg said in a statement: "Our investments are paying off, the core business is strong, and we're seeing velocity across the board." Chief Financial Officer Susan Li said the company is "still developing our budget and capacity plans for next year, but we expect to continue investing aggressively to support both our core business and emerging A.I. initiatives."

The broader industry is spending at a similar pace. Gartner forecasts worldwide information-technology spending will rise 15% next year to $7.3 trillion in 2027, from a projected $6.39 trillion this year. Data-center systems spending is expected to grow 55%. John-David Lovelock, distinguished VP analyst at Gartner, told the WSJ Leadership Institute: "We have the demand for AI-optimized servers going crazy. We have the supply not being able to keep up. We have AI servers doubling in price, and nobody's slowing down on orders."

Rival launches arrived the same day. Anthropic introduced Claude Duet, an early-test work agent joining OpenAI's dots, Meta's Muse, Google's Gemini Spark and agents from start-up Instinct, which raised $1 billion in its latest round. Duet sends users a morning briefing drawn from their email, calendar and business software, and it can connect to more than 6,000 outside apps. Users must approve every purchase; stock trading and money transfers between accounts remain blocked even with approval. Anthropic said it will show users what it plans to do before sending a message or making a payment. The agent is available by wait-list to Pro and Max subscribers in the United States at no added charge.

Why This Matters: Meta is committing well over $100 billion a year to a technology whose payoff remains unproven, and paying for much of that bet with debt, off-balance-sheet financing and a stock structure that keeps control concentrated in one person's hands. Free cash flow has turned negative, long-term debt has risen by more than $20 billion in a year, and a $25 billion penalty in New Mexico now sits alongside a $9.89 billion settlement charge weighing on earnings. The consequences reach beyond shareholders. Children's safety, consumer privacy and data-center financing terms affect people who never voted on the strategy. Public oversight of how dominant platforms finance and deploy AI, and whether they answer for harm, will shape who carries the risk when the boom cools.