Meta's AI spending spree lays bare its compute dilemma

Meta's AI spending cut free cash flow 91% to $784 million, sending shares down 9% in premarket trade

Meta can simultaneously advance its own artificial intelligence ambitions while renting out scarce computing capacity to boost returns, Chief Executive Mark Zuckerberg signalled on Wednesday. Investors, however, are not convinced.

The Facebook and Instagram owner is spending billions of dollars on AI infrastructure, including chips, servers, energy and data centres, leaving it with free cash flow of just $784 million in the second quarter, a 91% decline from a year earlier. The slump sent Meta shares down about 9% in premarket trading on Thursday.

Pressed by analysts on Meta's strategy, Zuckerberg described computing capacity as a scarce strategic asset that the company should retain and expand rather than sell for short-term gains.

He acknowledged, however, that Meta had received several offers from businesses willing to pay "a meaningful premium" to use its computing capacity for their own AI projects.

The tension highlights Meta's challenge in diversifying its revenue. Selling compute could ease pressure on cash flow but would also divert resources from its own efforts to develop AI models and services.

Having built its business on digital advertising through Facebook and Instagram, Meta is now competing with Microsoft, Alphabet and Amazon, which have established enterprise cloud businesses and are already generating significant AI-related revenue.

Microsoft demonstrated the strength of that model on Wednesday. Despite a 23% drop in free cash flow, the company exceeded expectations for Azure cloud growth and Copilot adoption, sending its shares up about 8%.

Selling intelligence over compute

"We believe that there will continue to be a significantly higher margin on selling intelligence rather than selling compute directly, but we think that there's a big opportunity to sell compute as well," Zuckerberg said.

He said Meta expects AI-powered personal assistants to become products used by billions of consumers, while business AI agents could eventually support customer service, sales and marketing.

Beyond broad references to subscription services and enterprise products, however, Zuckerberg provided few details on how those businesses would generate returns sufficient to justify Meta's massive AI spending.

Responding to J.P. Morgan analyst Doug Anmuth, Zuckerberg said Meta was deliberately investing ahead of demand because data centres require years to build before becoming operational.

"There is a lead time where we're investing in building out these data centres now. They come online at some point in the future. You obviously are not getting value out of them until they're online," he said.

Spending like a cloud giant

"Meta is spending like a hyperscaler without a hyperscaler's business model," said Josh Gilbert, APAC market analyst at eToro.

"Microsoft, Alphabet and Amazon can point their data centre dollars at cloud businesses that sell compute straight back out the door, but Meta doesn't have the same outlet, so every dollar of build-out leans on the ads business," he said.