Meta has entered into new agreements for AI computing with data center company Crusoe.

      Meta has entered into new agreements to access computing power from Crusoe, a data-center developer, further contributing to the extensive and expanding infrastructure that the company is assembling for its AI pursuits. These deals encompass two locations, one in Childress, Texas, and another in Warrenton, Missouri, providing Meta with approximately 1.6 gigawatts of capacity in total, as per a Bloomberg report citing sources familiar with the situation.

      This capacity figure is significant. A gigawatt reflects the approximate output of a sizable power plant, and having 1.6 gigawatts dedicated to a single company's computing requirements indicates a shift in AI ambitions from relying on chips to a focus on electricity. Meta is no longer merely purchasing servers; it is securing electrical capacity at a scale comparable to national grid planning, and doing so with multiple developers simultaneously.

      Crusoe is currently the key partner for this arrangement. Established in 2018, the company specializes in constructing and operating large data centers tailored for AI applications. It has already formed agreements with Oracle, Microsoft, and Alphabet's Google for three other sites.

      Including Meta in that lineup positions Crusoe among the developers that major AI buyers approach when they require capacity more quickly than they can establish it independently.

      For Meta, these agreements represent a component of a broader and more costly strategy. Similar to its Silicon Valley counterparts, the company has been striving to secure computing power wherever possible, with its most significant initiative being a nearly 4,000-acre campus in Louisiana aimed at providing up to five gigawatts of capacity. The Crusoe agreements serve as an incremental boost to this already significant expansion.

      This expansion is putting pressure on the surrounding resources. Data centers of this magnitude require power and water at a level that has begun to attract scrutiny from grid operators and local communities, and the competition for electricity is now a critical limitation for the industry, evident in operators racing to secure gigawatts across various continents. Meta's strategy has been to diversify its investments across different developers and locations rather than depending on a single site.

      The financial commitment involved is staggering and continues to grow. The largest cloud and AI companies are collectively investing hundreds of billions into infrastructure, with demand surpassing even the most ambitious forecasts. Some operators are even renting capacity from unconventional sources to keep pace.

      Meta's agreements with Crusoe are a small entry in this extensive financial ledger, yet they indicate a clear direction: secure power now, regardless of the cost, and focus on returns later.

      What Meta plans to utilize the 1.6 gigawatts for aligns with its familiar goals, including training and deploying the AI models and agents around which the company has reorganized itself. This capacity serves as a prerequisite for the rest of its strategy; without it, the models and products based on them cannot scale.

      Opting to lease capacity from developers like Crusoe, instead of building every site internally, is a strategic decision in itself. Constructing gigawatt-scale data centers requires years and considerable capital investment; contracting for capacity that others build and manage enables Meta to act more swiftly and spread construction risk, albeit at the cost of paying a margin to the developer. With demand outpacing supply, speed has become more valuable than ownership, and Crusoe offers exactly that.

      Neither Meta nor Crusoe has provided detailed comments on the agreements' terms, and the figures reported come from sources familiar with the arrangements rather than an official announcement. What these deals highlight is the trend: a company whose future hinges on AI is acquiring the electricity needed to power it in gigawatt quantities, from whoever can supply it most swiftly.