This article first appeared on GuruFocus.
Meta Platforms (NASDAQ:META), a technology company that develops artificial intelligence products and infrastructure, is exploring the possibility of creating more value by renting out portions of its computing network to external customers. CEO Mark Zuckerberg said demand for AI computing power is so strong that some external offers could be more attractive than using that power exclusively for Meta’s internal projects. However, Zuckerberg stressed that Meta is not necessarily building more infrastructure than necessary, noting that the company is currently using all of its available computing power and still has significant internal demand. Investors may view potential cloud expansion as an opportunity for Meta to generate revenue directly from data center, chip, and compute partnerships that support its AI operations.
Meta is developing several potential cloud products, including a service that will give developers access to AI models running on the company’s existing infrastructure. The structure could be similar to Bedrock, a service offered by Amazon Web Services, Amazon’s cloud computing business, where Meta operates the data centers and chips and charges developers a fee for its hosted models, including its own Muse Spark model. Zuckerberg did not confirm that Meta would host competing AI models, but said it makes sense to evaluate the possibility. Meta is also considering selling access to raw computing power in a model similar to cloud-computing company Coreweave, but Zuckerberg said long-term demand from customers could act as a backstop if Meta ends up owning infrastructure it doesn’t need for its own operations.
Zuckerberg also pointed to the strategy adopted by Elon Musk’s space company SpaceX (NASDAQ:SPCX). The company leases access to its Memphis data center to artificial intelligence developer Anthropic and has a separate agreement with Google, a technology company that provides computing services. He said Meta has received similar offers, so short-term computing contracts with large premiums may be worth considering. Meta has already moved to another form of paid AI access by announcing plans to offer Muse Spark 1.1 through an application programming interface and charge developers based on the number of tokens they use. While this approach could create additional revenue channels for Meta, Zuckerberg said the company continues to use its computing power significantly in-house.
