Meta is building plans to sell access to its AI compute power and models, according to a Bloomberg report surfaced by TechCrunch AI on Wednesday. The move would turn Meta’s massive data center buildout into a cloud business, putting it head to head with Amazon Web Services, Google Cloud, and Microsoft Azure. This is a real shift in strategy, and it says something important about where the AI race is actually being won.
What’s happening
Meta has poured billions into AI and the data centers that run it. Now it wants some of that money back faster. TechCrunch AI reports that the new effort, reportedly called Meta Compute, would work two ways:
- Raw compute for rent. Copying CoreWeave’s model, Meta would lease out spare capacity to other companies that need to train or run AI.
- Model access. Following AWS, Meta would let customers tap various AI models hosted on its infrastructure, including its recently launched closed-weight model, Muse Spark.
The initiative is led by head of infrastructure Santosh Janardhan, Meta Superintelligence Labs leader Daniel Gross, and president Dina Powell McCormick. It also confirms what Mark Zuckerberg floated in May, when he said a Meta cloud business was “definitely on the table.”
Why Meta is doing this now
Here’s what stands out. Unlike Google and OpenAI, Meta hasn’t seen big outside demand for its own AI models and services. It doesn’t even break out revenue from Meta AI or its Llama models in earnings, and executives mostly talk up internal corporate uses. In plain terms, Meta’s AI products aren’t yet a standalone money line.
But the spending is enormous. As of the end of the first quarter, Meta had committed $182.9 billion to AI infrastructure in the coming years, per TechCrunch AI. That includes huge projects in Louisiana and Ohio, with the Ohio site (which Zuckerberg said would be the size of Manhattan) expected to come online this year. Selling excess compute is a way to earn a return while the model business catches up.
The bigger signal
Meta isn’t first here. The company’s plan comes just weeks after SpaceX, through xAI, announced similar moves. In early May, SpaceX signed a deal with Anthropic to buy out all the compute at its Colossus 1 data center, and it has since signed leases with Google and Reflection AI.
When two of tech’s biggest players both pivot toward renting out compute, it tells you something: the winners of the AI race may not be the ones with the best models. They may be the ones who own the data centers. Compute is becoming the product.
The catch
That thesis only holds under two conditions. Demand for compute has to keep climbing, and data centers have to hold their value. Neither is guaranteed.
Skeptics have warned that the infrastructure buildout looks like a bubble leaning on chips that depreciate fast. Others question whether AI companies can pull in enough end-user revenue to justify trillion-dollar bets. If demand softens or chips age out quicker than expected, all that capacity turns from an asset into a liability.
What to watch
For anyone building on AI, more compute suppliers is good news. A Meta cloud business would add another serious option alongside AWS, Azure, Google, and CoreWeave, which usually means better availability and competitive pricing over time.
A few things worth tracking:
- Pricing. Whether Meta undercuts the big three to win share, or positions as a premium option tied to its models.
- Muse Spark. How Meta bundles its closed-weight model into the offering, and whether developers actually adopt it.
- Timing. The Ohio project coming online this year could be the trigger for a real launch.
Meta hasn’t confirmed the details, and TechCrunch AI notes the company was reached for comment. Read the full report at the original source for more.