Meta's Capital Expenditure Soars on the Back of AI Spending

Along with better-than-expected earnings, Meta on Wednesday reported record capital expenditures that hit $14.8 billion in Q4 of 2024 and $39.2 billion for the full year. The company also said that it expects to spend up to $65 billion on the same item in 2025 as it plans to expand AI infrastructure and staffing. If realized the spending increase would be as big as 130 percent in just two years.

Microsoft and Amazon earlier in the month made similar investment projections for the current year, with Microsoft planning to put $80 billion into data centers and Amazon saying that it would surpass its initially estimated spending of $75 billion for the type of infrastructure.

While Meta is best-known for its forays into virtual reality and the metaverse, the company has a popular AI product, Llama, that follows a complete open-source approach unlike many industry leaders. This means that Meta is not earning money from Llama through traditional licenses or subscriptions, even though the company introduced commercial licenses that allows other enterprises to integrate it into their products in use cases that are deemed to broad under the open-source license.

However, Meta is hoping that its different approach will be a benefit down the line and help them in an area that suddenly became important in the AI sphere with the launch of Chinese AI DeepSeek-R1, which claims to have cut costs significantly. Open-source means that Llama is continuously being improved by outsiders and that Facebook can benefit from their expertise at a fraction of the cost of hiring them. Meta could also draw these developers into their sphere by offering Llama open-source and hopes that it can recruit cutting-edge talent later. By offering open-source, the company is also hoping to become a standard in many fields, which will then help them with its future AI forays.