Meta Platforms Inc. has inked deals to acquire roughly 1.6 gigawatts of computing capacity from data center developer Crusoe, according to sources cited by Bloomberg. The capacity will be spread across two data centers in Texas and Missouri. The agreements significantly bolster the infrastructure Meta needs to support its AI ambitions.
The move underscores the intensifying race among tech giants to secure vast amounts of energy-hungry computing power for training and running AI models. By partnering with Crusoe, Meta taps into a developer known for using natural gas to power on-site data centers, a strategy that aims to reduce grid strain. This comes as Meta and peers like Microsoft and Google compete for limited data center space and energy resources.
The 1.6 GW figure represents a massive scale, far exceeding the capacity of typical hyperscale data centers. For context, large AI training clusters often operate at several hundred megawatts. Bloomberg's sources did not disclose the financial terms of the contracts or the precise timeline for when the capacity will come online.
These agreements could help Meta reduce its reliance on public cloud providers for AI workloads, giving it more control over costs and performance. However, the environmental impact remains a question, as Crusoe's natural gas approach, while efficient, still produces emissions. Regulatory scrutiny of data center energy use is also intensifying.
Neither Meta nor Crusoe has publicly commented on the deal. The arrangement highlights how AI's infrastructure demands are reshaping the data center industry and energy markets.