(Reuters) - Meta Platforms fourth-quarter revenue beat Wall Street expectations on Wednesday but the company predicted sales in the current first quarter may not meet forecasts, sending mixed signals about how bets on pricey artificial intelligence-powered tools are paying off.

The Facebook and Instagram parent company expects first quarter revenue between $39.5 billion and $41.8 billion, compared with analysts' average estimate of $41.72 billion, according to data compiled by LSEG.

Meta shares were flat after the market closed but rose 4.5% as CEO Mark Zuckerberg spoke optimistically about Meta's AI initiatives.

"Meta AI is already used by more people than any other assistant, and once a service reaches that kind of scale, it usually develops a durable, long term advantage," he said on a conference call after reporting results.

The forecast raised fresh questions about Meta's capital spending. The company relies on its core social media ads business to cover the costs associated with its AI ambitions and investments in "metaverse" technologies like smart glasses and augmented reality systems.

Last week, Zuckerberg announced that Meta plans capital expenditure of as much as $65 billion in 2025 to expand its AI infrastructure, while also increasing hiring for AI roles.

On Wednesday, Meta said it expected total expenses for 2025 to be in the range of $114 billion to $119 billion, up from a total of $95 billion in 2024.

"Meta's gangbusters Q4 results clearly demonstrate that ad revenues remain the company's lifeblood. That said, the biggest question heading into 2025 isn't about today's earnings-it's about whether Mark Zuckerberg's $60-65 billion AI infrastructure bet will pay off," said Jeremy Goldman, principal analyst at eMarketer.

Family daily active people (DAP), a metric Meta uses to track unique users who open any one of its apps in a day, rose about 5% from a year earlier to 3.35 billion.

Meta's results come after Chinese startup DeepSeek's launch of its latest AI models triggered a selloff in global tech stocks on Monday on concerns about rising AI costs in the U.S.

DeepSeek has said its models either match or outperform top U.S. rivals at a fraction of the cost, including Meta's own Llama models, challenging the prevailing view that scaling AI requires vast computing power and investment.

The breakthrough could heighten scrutiny from investors worried about the company's heavy spending on AI, though it may also benefit Meta if it successfully brings down the cost of building and supporting the models.

Meta - among the top buyers of Nvidia's sought-after AI chips - aims to end the year with over 1.3 million graphics processors (GPUs) and bring about 1 gigawatt of computing power online, Zuckerberg said on Friday in a Facebook post outlining the company's 2025 spending goals.

He said this month that Meta would lay off 5% of its "lowest performers" and warned employees about more such job cuts this year to raise performance.