Facebook: Meta Plans To Save $4bn On Data Centers And Create New AI Facilities Faster and Cheaper

Feb 08, 2023 | Posted by Abdul-Rahman Oladimeji

In a cost-cutting effort, Meta claims to have rebuilt its data centers to be less expensive and quicker to build. Meta canceled or paused data centers worldwide and redesigned them for additional artificial intelligence demands late last year. Facebook's Danish data center development and Texas, Alabama, and Idaho projects were halted. 

Before the dramatic turn, Meta asserted that its larger cuts would not affect data center spend since its main company was under assault, its metaverse attempts failed, and the world economy weakened. Meta detailed the financial implications of the new designs in its last quarterly report. Meta's CFO Susan Li claimed "the write-down of some data center assets" increased last quarter's cost of revenue by 31%. Capex for the quarter was $9.2 billion, "fueled by expenditures in servers, data centers, and network infrastructure." 

However, it now expects a reduced capex outlay. Instead of $34-37bn, it estimates capex (excluding data centers) to be $30-33bn. Li said the decreased expectation reflects their updated intentions for lower data center construction investment in 2023 as they switch to a more cost-efficient data center design that can accommodate AI and non-AI workloads. She said that that will give them more options as they learn about AI demand. The new data center design should be cheaper and faster to build. 

The business was still hazy about the new designs, and it is unclear how much cost reductions came from the designs rather than a decreased data center space. Li said that the firm is streamlining its data center development process to have a new phased strategy that allows them to design base plans with less initial capacity and capital outlay but swiftly stretch up future capacity if needed. That implies that the corporation will construct modularly and grow more slowly than its existing method, which involved building large campuses with massive data halls.

0 Comments