Mark Zuckerberg spent big on artificial intelligence in 2025, but his New Year’s resolution might be to tighten the belt at Meta Platforms.

The costs of massive data centers and top-tier AI researchers are piling up, forcing Meta to get creative in its quest for capital, securing $27 billion of funding for its Hyperion supercomputer in Wall Street’sbiggest-ever private credit dealin October. Investor sentiment reached a tipping point after Zuckerberg shared plans to ramp up AI spending even further on company’s third-quarter earnings call. Through Tuesday, Meta’s stock has fallen 15.8% from the Aug. 12 record close of $790.00.

The pressure is on for Zuckerberg to deliver a return on his spending spree and show that the company can manage costs. Although AI technology has helped Meta grow its core advertising business, expenses grew faster than revenues in the third quarter, and some analysts are concerned the trend could continue as depreciation expenses from data centers start hitting the income statement.

Another concern from BNP analyst Stefan Slowinski is that Meta’s business model is highly dependent on advertising revenue, unlike other Big Tech peers such as Alphabet and Amazon.com, which have cloud-computing businesses. When Amazon and Google build a data center, they can rent out the excess capacity to external customers, something that Meta can’t do, Slowinski told MarketWatch last month.

In a December note, Morgan Stanley’s Brian Nowak said that more detailed guidance on Meta’s operating expenses could help change the narrative. He suggested that the upcoming January earnings report could be a “clearing event.”

Recent reports of dramatic budget cuts in the metaverse department, which focuses on virtual worlds, have given investors newfound optimism for Meta. The metaverse sits under Meta’s Reality Labs division, which incurred a $4.4 billion loss last quarter.

Meta went through a noteworthy cost-cutting initiative in 2023, after Zuckerberg declared a “year of efficiency” and slashed spending on the metaverse. If Meta were to go through with more cuts next year, it could have more resources to focus on projects with a higher return on investment, such as its advertising algorithm and its Ray-Ban AI glasses.

“Within our overall Reality Labs portfolio we are shifting some of our investment from metaverse toward AI glasses and wearables given the momentum there,” a company spokesperson told MarketWatch. “We aren’t planning any broader changes than that.”

Investors will want to see Meta exploring new ways of AI monetization through the company’s Meta AI assistant, Messenger Bots and wearables. The company reported on its last earnings call that Meta AI has over one billion monthly active users, which reflects the company’s distribution power across different social-media platforms. However, the statistics around monthly active users don’t show how sticky engagement is, or whether it’s comparable to ChatGPT or Gemini, Slowinski said.

Meta Superintelligence Labs, the newly formed AI unit led by Alexandr Wang, is also under pressure to perform, as Meta’s latest Llama 4 model underwhelmed users earlier this year. The Wall Street Journal reported that the company is developing an image and video model code-named Mango and a text model code-named Avocado for release next spring.

Bloomberg has also reported that Meta could be developing a closed-source model, meaning that the technology will no longer be accessible to the public to use for free.

Another monetization method for Meta would be to distribute its models through cloud providers like Amazon and Microsoft’s, according to Nowak. In this scenario, customers would pay to rent access to Meta’s AI through a cloud platform instead of hosting it themselves.

Nowak believes the Superintelligence unit could actually be the most overlooked part of Meta right now, with the potential to send the stock to $1,000, which is about 50% above Tuesday’s closing price of $664.94.