(July 30): Meta can accelerate its AI ambitions while also lending out its scarce computing power to boost profits, CEO Mark Zuckerberg suggested on Wednesday. The problem is that investors aren’t buying it.
Instagram’s owner has spent billions of dollars building out computing (chips, servers, energy, data centers that power AI), and the free cash flow needed to run and grow the business was just US$784 million in the second quarter. The stock fell more than 9% on Thursday, a 91% year-over-year decline.
Pressed by analysts for details on Meta’s plans, Mr. Zuckerberg framed computing as a rare strategic asset that the company should retain and build around, rather than simply selling it for short-term gains.
However, he acknowledged that the company had received a number of offers for its computing power from companies who wanted to roll out their own AI plans at a “significant premium” over what the company had invested in building it.
This tension is at the heart of Meta’s challenge to diversify its revenue sources. Renting the compute could ease pressure on Meta’s cash flow, but it would also divert scarce resources from the company’s push to build AI models and services.
The company, which built wealth-selling ads on Facebook and Instagram, is now competing against larger rivals such as Microsoft, Alphabet and Amazon, which have deep ties to companies that were an early and lucrative market for AI.
Microsoft on Wednesday showed how its bet on AI is paying off, even as free cash flow fell 23%. The Windows maker smashed growth expectations for its Azure cloud division and Copilot assistant, sending its stock up more than 13% thanks to its vast enterprise customer base and early AI buildouts.
Spend like a cloud giant
“We believe that selling intelligence will continue to yield significantly higher margins than selling compute directly, but we also clearly see a significant opportunity in selling compute,” he said, painting a picture of what Meta hopes to build with its significant investment in AI.
Zuckerberg argued that AI-powered personal assistants could become mass-market products used by billions of consumers, while business agents could eventually help companies handle customer service, sales and marketing.
But beyond broad references to subscription and enterprise services, he offered few specifics about how those businesses justify Meta’s massive AI spending.
In response to a question from JPMorgan analyst Doug Anmuth, who pointed out that while Meta sells compute, it also buys capacity from third parties, Zuckerberg said Meta is intentionally investing ahead of demand.
“There’s a lead time right now where you’re investing in building these data centers, and they’ll come online at some point in the future. Obviously, you’re not getting any value until they’re online,” he said.
“Meta is spending like a hyperscaler without having a hyperscaler business model,” said Josh Gilbert, principal APAC analyst at online investment platform eToro, referring to the big cloud companies.
“While Microsoft, Alphabet, and Amazon can redirect their data center funding to cloud businesses that sell compute outright, the meta doesn’t have the same outlet, so every dollar built relies on the advertising business.”
horror of the metaverse
Meta’s massive investment push in AI has some investors worried. They point out similarities with the company’s expensive pivot into the Metaverse. The bet didn’t turn into a huge source of profit, but it racked up tens of billions of dollars in losses.
The drop in free cash flow in the April-June period was the biggest since late 2022, when bets on the Metaverse drew similar investor scrutiny.
Still, Meta raised the lower end of its capital spending forecast by $5 billion, raising the range to $130 billion to $145 billion.
This is similar to Alphabet, which raised its own forecast by $15 billion last week, reported its first-ever cash burn, and was punished with a 7% share price drop.
Meta’s Chief Financial Officer Susan Lee said in a post-earnings conference call that the spending was justified because the industry has been “historically under-built” for AI demands and existing capabilities are “extremely valuable.”
He said production capacity will remain tight “for the foreseeable future,” creating opportunities for Meta to generate profits through products, enterprise services and computing sales.
But in response to a question from Wells Fargo analyst Ken Gawlerski about whether Meta plans to keep all of its computing power in-house through 2027, Lee ended the call just as Zuckerberg began. Meta asserted that it could find beneficial uses for the additional computing both internally and externally, but did not indicate where the ultimate benefit would be greatest.
“The earnings call was a lot like a good old-fashioned brainstorming session,” said Mark Schmulik, an analyst at Bernstein.
Uploaded by Lam Sen Fat
