Hajra Naz
Meta has invested billions of dollars in artificial intelligence. It has built huge data centers and expanded its AI computing infrastructure. Now, the company wants to earn money from those investments.
A recent Bloomberg report says Meta is planning a cloud infrastructure business. The service would let outside customers rent AI computing power. Businesses could also access Meta’s AI models. If the plan moves ahead, Meta will compete with major cloud providers like Amazon Web Services (AWS), Google Cloud, and Microsoft Azure.
The move comes just weeks after SpaceX, through its AI company xAI, revealed a similar strategy. In May, SpaceX reached an agreement with Anthropic to utilize the full computing capacity of its Colossus 1 data center. The company has since signed additional compute agreements with Google and Reflection AI. Meta’s reported plans suggest that owning large-scale AI infrastructure could become just as valuable as building advanced AI models.
Many industry analysts believe the companies that control powerful data centers may be in the strongest position as AI demand continues to grow. However, not everyone is convinced that the current pace of AI infrastructure spending is sustainable.
Read More: The Next Big Computing Revolution Is Coming—And It’s Not AI
Some experts argue that the rush to build AI data centers could be creating an infrastructure bubble. They point to the enormous costs of AI chips, which lose value over time, and question whether AI companies will generate enough revenue from customers to justify investments that now total hundreds of billions—or even trillions—of dollars.
Despite those concerns, Meta has continued to increase its AI spending.
By the end of Q1, it had already booked around $182.9 billion worth of spending on AI infrastructure projects over the next few years. Involved in this investment are large data center projects in the US, such as those planned for Louisiana and Ohio. Expected to be of similar scale to Manhattan, CEO Mark Zuckerberg once said it was going to be opened in Ohio this year.
Meta has spent heavily on AI infrastructure, but its AI products have not made the same commercial impact as some competitors to date.
Compared to Google or OpenAI, Meta has not shared any revenue statistics for Meta AI or the open-weight AI models from its Llama family. Meta’s executives in earnings calls have mostly focused on how AI impacts the function of its products and services as well as its internal operations, rather than on the fact that it is a major business in itself. It has caused many observers to conclude that the AI services of the company are not yet producing a lot of direct income.
Read More: Smarter Than ChatGPT: Why AI Agents Are the Future of Financial Services
Selling computing power could help Meta recover part of its massive infrastructure investment.
Bloomberg reported that the company is considering a model similar to CoreWeave, offering customers access to raw GPU compute capacity. At the same time, Meta is also evaluating a marketplace similar to AWS, where businesses could access multiple AI models running on Meta’s infrastructure, including its recently introduced closed-weight model, Muse Spark.
The initiative is reportedly being developed under a new division called Meta Compute. The effort is said to be led by Meta’s head of infrastructure, Santosh Janardhan, alongside Meta Superintelligence Labs leader Daniel Gross and company president Dina Powell McCormick.
The report also aligns with comments Zuckerberg made in May, when he said that creating a cloud computing business was “definitely on the table.” At the time, he suggested that renting out Meta’s computing infrastructure could help generate returns on the company’s enormous investment in its long-term vision of building AI superintelligence.
