Meta expands full-stack AI strategy as Zuckerberg weighs new revenue beyond ads

Highlights

Meta approved $100–200 billion AI infrastructure buildout, with up to $145 billion in 2024 capital expenditures, despite lack of clear performance evidence.
Meta launched Muse Wonder frontier-class model and is developing a cloud business to sell AI compute, aiming to hedge massive infrastructure costs and diversify revenue.
Meta's Muse AI agents could shift revenue mix beyond ads—currently 98% of income—with new models including $20–$100 subscriptions and commerce transaction fees.

AI spending plan and platform control

As reported by Fortune, Zuckerberg says Meta approved an AI buildout that he estimated at between $100 billion and $200 billion before there was proof the effort would succeed. On the Oct. 12 episode of Cheeky Pint, he and chief AI officer Alexandr Wang describe that scale of decision-making as a long-term bet made ahead of clear performance evidence.

The same day, Meta launches Muse Wonder, described as the first frontier-class model from the superintelligence lab led by Wang. Meta has guided for as much as $145 billion in capital expenditures this year, with much of the spending directed toward AI infrastructure.

Zuckerberg says the company is pursuing a full-stack approach that includes data centers, chips and low-level systems. He links that strategy to Meta’s long-running dependence on Apple and Google mobile platforms, a relationship marked by restrictions on app distribution, data access and tracking, including Apple’s 2021 privacy changes that Meta later says would cost it about $10 billion in 2022 alone.

Meta also carries the cost of that expansion without a traditional cloud business to offset infrastructure spending. Zuckerberg says a one-gigawatt AI data center costs $50 billion to $60 billion to build and about $15 billion a year to operate, while the company’s Hyperion campus in Louisiana is designed to scale to five gigawatts.

That model may broaden over time. Bloomberg reported in July that Meta is building a cloud business to sell AI computing power, and Zuckerberg later confirms on the company’s earnings call that cloud plans are underway, presenting the compute capacity as a hedge if demand for AI infrastructure remains strong across the wider market.

Agent commerce could reshape Meta’s revenue mix

Zuckerberg and Wang say Meta wants Muse Wonder and Muse Wonder Full to power Muse, a personal AI agent designed to act on behalf of users. Wang describes the product as a digital team of specialists, while Zuckerberg says he already uses it through Instacart to order ingredients for baking projects, increasing his own usage from occasional orders to once or twice a week.

That push into agentic commerce raises questions about the future of Meta’s ad business. Advertising accounts for about 98% of Meta’s revenue in the latest quarter, and eMarketer projects the company’s worldwide ad revenue reaches $243.46 billion this year, overtaking Google’s for the first time.

If AI agents compare prices, review delivery options and make purchase decisions directly, the role of traditional digital advertising could weaken. Zuckerberg argues businesses are instead asking to integrate more deeply with agents, and says broader automation could expand the overall commerce market rather than simply displace existing ad activity.

He also points to revenue streams beyond advertising. Zuckerberg says subscriptions are likely to become important, with Muse offered free and paid tiers at $20 and $100 a month, and he suggests Meta could also take a small share of savings or gains generated for users through commerce transactions. On the company’s July earnings call, he says business agents could operate much like the ad system by charging when Meta delivers measurable results.

What remains unresolved is whether those agents will erode the ad impressions that support Meta’s current business model. For now, that risk is not drawing significant pressure from Wall Street, even as the company spends aggressively to build the infrastructure behind its AI strategy.

Our earlier article on U.S. AI data center expansion explained how the infrastructure boom is running into higher financing costs, construction delays, and power constraints that make many planned projects look less credible. We noted that tightening capital markets, cost overruns, and growing local and regulatory pushback are raising investor concerns about whether developers can deliver new compute capacity on time and at sustainable returns.

This material may contain third-party opinions, none of the data and information on this webpage constitutes investment advice according to our Disclaimer. While we adhere to strict Editorial Integrity, this post may contain references to products from our partners.