Meta Stock Has Slumped on AI Spending. The Bigger Payoff May Be Ahead
Meta keeps making the kind of investments that make investors nervous, and so far, they keep paying off. The stock is down 6% this month on rising AI infrastructure costs, but three distinct AI bets are taking shape while the market remains focused on just one.
Selling intelligence is just the start
Advertising still pays the bills, recently generating $60.8B in quarterly revenue for Meta. But that business also gives the company a major advantage in AI, with more than 2 billion people opening Instagram each day.
Meta plans to tap that reach with “Hatch,” a consumer AI agent built directly into Instagram, removing the need for a separate app or account.
A tiered subscription model, reportedly priced as high as $199.99 a month, could turn that audience into a new stream of recurring revenue.
Hatch is also being trained to work with services like DoorDash, Etsy, and Microsoft’s Outlook, while Meta’s next AI model, “Watermelon,” is slated for October.
Meta's hidden compute business
The second and more overlooked bet involves physical infrastructure. Meta is targeting 14 gigawatts of total compute capacity by 2027 while producing its own Iris chips.
With no public cloud business of its own, Meta could eventually sell some of that capacity into a market where supply is already tight.
Evercore ISI analyst Mark Mahaney estimates that selling just 0.5 to 1 gigawatt could bring in $11B to $22B in annual gross revenue and add up to $4.32 per share in earnings.
That would use only 7% of Meta’s planned capacity at the high end. Zuckerberg says buyers are already willing to pay a premium, leaving Meta with what Mahaney calls a “call option” on an entirely new revenue stream.
When AI goes rogue
None of this is riskless. Net profit fell from $18.34B to $15.85B year-over-year despite revenue climbing from $47.52B to $60.8B, a gap that reflects the scale of current AI investment.
Corporate insider sentiment is currently negative, with insiders net sellers over the recent quarter. And Meta faces a threat that's genuinely new territory for any tech company.
Meta’s AI push also comes with a growing security risk. During testing by Israeli startup Irregular, a pre-release Meta model exploited a vulnerability in a real company’s website after a testing misconfiguration gave it internet access.
Meta says the incident was isolated and did not involve a sophisticated sandbox escape, though similar breaches involving OpenAI and Anthropic models have raised broader concerns about containing increasingly capable AI systems.
For Meta, those concerns could add regulatory and reputational pressure as it prepares AI agents to handle more sensitive tasks for users.
"We are handling A.I. with our bare hands, and we don't know how to contain it, let alone how to safely test it."
Katie Moussouris, Luta Security
Bank of America maintains a Buy rating on Meta, while the broader analyst consensus remains at Strong Buy with roughly 38% upside from recent levels. Evercore’s Mark Mahaney sees that upside topping 50% if Meta starts monetizing its compute capacity.
Yet at just 17 times forward earnings, near a three-year valuation low, investors are largely paying for the advertising machine. Hatch and a potential compute business remain upside the market has yet to fully price in.