Meta has revealed plans to build its own Cloud Computing business to rent out computing power and AI models, creating new revenue streams and offsetting the massive investments it has made in AI infrastructure.
According to Bloomberg, Meta Platforms is developing plans to establish its own Cloud Infrastructure business, aiming to sell AI computing power and AI model access services, directly competing with major cloud providers like Amazon Web Services (AWS), Microsoft Azure, and Google Cloud.
Meta has recently invested heavily in data centers and costly infrastructure to support its AI development plans. Currently, with excess unused compute power available, Meta intends to monetize it by renting it out to external customers.
One plan under consideration is allowing customers to access various AI models running on Meta’s own infrastructure, similar to Amazon’s AWS Bedrock service.
Under this approach, Meta would manage the data centers and processing chips running AI models, including its own Muse Spark model, and charge developers for accessing and using these models.
Meta is also considering offering customers raw compute power rentals, similar to the business model used by Neocloud providers like CoreWeave.
All these new business developments fall under an internal project named “Meta Compute,” created to develop and manage the company's AI infrastructure, led by key executives including:
However, a Meta spokesperson declined to comment to Bloomberg, stating that the plans are still in development and the company’s strategy may change in the future.
Following the news, Meta’s stock rose 9.3% to $615.55 in morning trading on Wednesday in New York, marking the largest intraday gain since April and closing nearly 9% higher.
Conversely, shares of CoreWeave fell as much as 14%, while Nebius Group NV, a Dutch AI data center provider listed on the New York Stock Exchange, dropped as much as 17%.
Meta has elevated the development of AI Superintelligence—AI exceeding human capabilities—to one of its top priorities, announcing investments of hundreds of billions of U.S. dollars to build data centers, AI infrastructure, and procure expensive processing chips deemed essential to achieving this goal.
However, such massive investments have raised investor concerns about how Meta will generate returns from these expenditures, especially as demand for compute power from major AI developers continues to grow without signs of slowing.
Previously, Meta signed agreements to source compute power from companies like CoreWeave, Alphabet’s Google, and Oracle to meet increasing AI usage demands.
In recent quarters, Meta and other major tech firms have announced huge investments to reserve data center capacity for their own use, significantly increasing AI infrastructure spending across the industry.
Thus, launching its own cloud business could be another way for Meta to monetize these investments.
For decades, leading cloud providers including AWS, Microsoft Azure, and Google Cloud have invested in platforms allowing customers to rent compute power, storage, and software over the internet, creating multi-billion-dollar quarterly businesses.
As AI demand surges, these providers have expanded from general server rentals to offering specialized chips and compute power tailored for training and running AI models.
However, this business is highly complex, requiring massive data center investments as well as strong software platforms, enterprise sales teams, and customer support.
Despite the complexities, Meta CEO Mark Zuckerberg has repeatedly signaled openness to monetizing excess AI infrastructure to investors.
Beyond selling compute power, Meta may offer API services where customers pay based on AI usage, typically measured by the number of tokens or data processed during AI operations.
Mark Zuckerberg said during a shareholders' meeting in May, “This is a serious option we are considering.”
He explained that almost every week, outside companies inquire whether Meta is interested in offering API services or selling compute resources at prices above the company’s invested costs.
He added, “We haven’t done this yet because we still believe the company itself needs that compute power.”
However, he noted that if Meta ever builds infrastructure exceeding its actual needs, monetizing excess resources would be an option the company is ready to consider. This idea also underpins Meta's confidence in continuing to invest heavily in AI infrastructure.
Amid intensifying AI competition, Mark Zuckerberg has emphasized that he believes compute power will be the scarcest resource in the industry’s future, so Meta should accumulate as much of it as possible before deciding how to utilize it.
Follow the Facebook page: Thairath Money at the link -https://www.facebook.com/ThairathMoney