Meta has invested billions in artificial intelligence development and the expansion of its data center operations. Now, the tech giant appears poised to leverage those data centers for a more immediate revenue stream.
On Wednesday, Bloomberg revealed that Meta is working on plans to create a cloud infrastructure service, which would provide access to both AI computing resources and models. This strategic pivot puts Meta in direct competition with major cloud service providers such as Amazon Web Services, Google Cloud, and Microsoft Azure.
This initiative follows closely on the heels of SpaceX's recent announcement through its xAI division about similar cloud computing plans. Earlier in May, SpaceX secured a deal with Anthropic to utilize all computing capacity at its Colossus 1 data center and has followed that with additional agreements with Google and Reflection AI. Meta's decision to adopt a similar approach signals a shift in the AI landscape, suggesting that the companies that ultimately succeed may be those that own the infrastructure, rather than merely providing superior models or services.
The potential success of this strategy hinges on sustained demand for computing power and the viability of data centers. Some analysts caution that the rush to expand AI infrastructure risks creating an economic bubble, given concerns over the rapid depreciation of hardware components. Furthermore, there are doubts about whether AI firms can generate sufficient revenue from end-users to justify their massive investments.
Despite these apprehensions, Meta remains committed to its AI infrastructure development, with plans to allocate $182.9 billion in the coming years. This includes ambitious ongoing projects in Louisiana and Ohio, with the Ohio facility expected to be about the size of Manhattan and operational within the year.
Unlike its competitors, such as Google and OpenAI, Meta has not yet experienced substantial demand for its own AI products and services. The company does not separately disclose revenue from its AI initiatives, including its Llama family of open-weight models, and public comments from executives have primarily focused on internal applications of AI. This suggests that Meta's AI projects may not currently contribute significantly to standalone revenue.
To recoup some of its significant investments, Meta may adopt a business model similar to CoreWeave by offering access to “raw” computing resources, according to Bloomberg. Additionally, reports indicate that Meta is exploring the possibility of selling access to various AI models, including its recently launched closed-weight model, Muse Spark, through its cloud infrastructure.
This upcoming venture is expected to be part of a new initiative referred to as Meta Compute, overseen by infrastructure head Santosh Janardhan, Meta Superintelligence Labs leader Daniel Gross, and president Dina Powell McCormick.
This aligns with Zuckerberg's statements from May, confirming that a cloud computing operation for Meta is “definitely on the table” as a means to generate returns from its substantial investments aimed at developing AI "superintelligence."
TechCrunch has reached out to Meta for further commentary.
