Meta shares jumped nearly 10% Wednesday following a report the company is planning to sell excess computing power, allowing it to recover some of the billions of dollars it has sunk into AI.
The Menlo Park, Calif.-based tech giant, which has rushed to secure pricey data centers and chips, is building a new cloud business to sell access to its AI models and compute, according to Bloomberg.
It’s welcome news for investors, who have grown anxious over whether Meta will be able to deliver returns on the hundreds of billions of dollars it has spent to build up a trove of coveted computing power, with a goal of developing “superintelligence.”
Meta declined to comment.
The new cloud business would allow Meta to generate revenue on any leftover capacity, while setting it up to compete with industry leaders like Amazon, Microsoft, Google, CoreWeave and SpaceX.
Meta is debating whether the cloud business should be structured to sell access to its own AI models, or to raw computing power itself, according to the report, which noted that plans could change.
If it decides to sell access to AI models on its own infrastructure, it would be taking a similar approach to Amazon – running the data centers and chips that power the bots and then charging customers fees to access them.
Elon Musk’s SpaceX – which took over his artificial intelligence firm xAI in February – has adopted a similar approach, striking lucrative rental deals with Anthropic and Google for access to its huge Memphis data center.
Anthropic agreed to pay $1.25 billion a month, while Google signed off on a $920 million monthly fee.
Meta could alternatively choose to sell access to its computing capacity, similar to CoreWeave’s business model.
OpenAI kicked off the race to amass large amounts of computing capacity in 2022 with the launch of its ChatGPT bot, as developers recognized that there was a limited amount of power despite skyrocketing demand.
In April, shares in Meta slid after the company raised its spending forecast to $145 billion amid mounting fears that AI stocks are overvalued, similar to the “dot-com bubble” of the early 2000s.
Meta CEO Mark Zuckerberg has repeatedly insisted that it’s crucial for the company to build up as much computing capacity as possible and consider its use later, since supply is limited – but in May, he signaled an openness to selling excess power.
“It’s definitely on the table,” Zuckerberg said at the annual shareholder meeting. “Almost every week there are different companies that come to us from the outside asking us to both stand up an API service or asking if we have compute that they could buy from us at some premium to what we’ve bought it at.”
“We haven’t done that yet because we think we have a use for the compute,” he added. “But obviously if we get to a point where we feel that we have overbuilt, then that is an option that we have, and that is partially what gives us confidence in investing in building this out.”
Last summer, Meta paid a whopping $15 billion to hire AI brainiac Alexandr Wang and take a 49% stake in his startup, Scale AI.
The company released its first AI model under Wang’s lead in April – though the model, called “Muse Spark,” did not live up to hopes for a state-of-the-art bot.
Wang has defended the model, saying it should serve as an “appetizer” while Meta is “cooking” up the main course.
