Mark Zuckerberg, CEO of Meta Platforms Inc., arrives for the Meta Connect event in Menlo Park, California, on Sept. 25, 2024.
David Paul Morris | Bloomberg | Getty Images
Meta has been so quick to build out its massive data center and computing infrastructure for artificial intelligence projects that CEO Mark Zuckerberg is even a bit surprised.
In a call with analysts on Wednesday after Meta’s third-quarter earnings report, Zuckerberg explained to investors how Meta’s rising costs for the year are tied to the speed at which employees are able to get data centers, servers and chips for AI up and running.
“Going into the year, we had a range for what we thought we could potentially do, and we’ve been able to do more than we’ve kind of hoped and expected at the beginning of the year,” Zuckerberg said.
It also means investors have to buckle up for higher expenses. Meta raised the low end of its capital expenditures guidance for 2024 to $38 billion from $37 billion. The top end is still $40 billion.
“It’s actually something that I’m quite happy that the team is executing well on,” Zuckerberg said. “That execution makes me somewhat more optimistic that we’re going to be able to keep on building this out at a good pace.”
Meta added that the expenditures, which include purchases of billions of dollars worth of Nvidia’s graphics processing units, will grow significantly in 2025.
Meta shares dipped in extended trading on Wednesday despite the company beating on earnings and revenue. Weaker-than-expected user growth was part of the concern, along with rising costs.
On the earnings call, Barclays analyst Ross Sandler asked Zuckerberg how quickly Meta can build the immense computing infrastructure needed to reach its goals around generative AI, given possible barriers like energy requirements and the time needed to develop its own custom AI-specific chips.
Zuckerberg responded by complimenting Meta’s infrastructure team, which he said was “executing quite well” in building out more computing capacity for various AI projects like the Llama family of large language models.
Wall Street has grown concerned that tech giants like Meta and Alphabet are spending too much on infrastructure without seeing immediate returns. It’s a theme Zuckerberg acknowledged in an interview in July with Bloomberg, telling Emily Chang that there’s a chance that companies are “overbuilding now.” However, the risks of underinvesting are too great, he said.
“The formula around building out the infrastructure is maybe not what investors want to hear in the near term, that we’re growing that,” Zuckerberg said on Wednesday. “But, I just think that the opportunities here are really big, we’re going to continue investing significantly in this and I’m proud of the teams that are doing great work to stand up a large amount of capacity so that way we can deliver world-class models and world-class products.”
It’s not the only place where investors have to stomach hefty expenses.
Meta’s Reality Labs unit, home of metaverse technologies, posted an operating loss of $4.4 billion in the third quarter. The company said it expects “2024 operating losses to increase meaningfully year-over-year due to our ongoing product development efforts and investments to further scale our ecosystem.”