Meta Stock Surges 10% As Company Considers Selling Excess AI Compute Capacity

Shares of Mark Zuckerberg's Meta jumped 10% on Wednesday after reports emerged that the tech giant was building a cloud business to sell excess AI computing capacity.

At 1:21 PM ET, Meta's stock was 9.94% higher at $619.29, according to Nasdaq data.

Stocks of neocloud firms CoreWeave and Nebius Group dropped about 12% each following the news, CNBC reported.

Meta's move to sell its excess computing power to outside customers will help offset the billions of dollars worth of investments it has made in AI.

With this, the Zuckerberg-owned company will compete with industry leaders like Microsoft Azure, Amazon Web Services and Google Cloud, as per Bloomberg.  

One potential plan by the tech giant pertains to selling access to various artificial intelligence models that are hosted on its existing AI infrastructure, people familiar with the matter told Bloomberg.

The approach would be similar to AWS's Bedrock offering. Meta would operate data centres as well as the chips that power the models, and charge developers to access them.

The firm is also looking at selling access to “raw” computing capacity, which could be similar to neocloud businesses such as CoreWeave.

These new business lines will be part of Meta Compute, an internal initiative to create and manage the firm's AI infrastructure efforts. Meta Compute is led by company president Dina Powell McCormick, Santosh Janardhan, the firm's head of infrastructure and Daniel Gross, a leader who is part of the Superintelligence Labs AI unit.

Meta is among the several companies that have been racing to secure computing power since the AI boom started in 2022. With its move to sell excess compute capacity, Meta will follow the lead of Elon Musk's SpaceX, which started doing the same early this year. SpaceX has inked deals with Google and Anthropic.

The company told investors in April that it aims to spend around $145 billion on capex this year as it continues to secure the graphics processing units required to train AI models and run large workloads. Another focus of the company would be to develop AI data centres.

Meta has floundered when it comes to finding its footing in the AI industry, even after it spent $14 billion last year to bring in Alexandr Wang from Scale AI. In April, the tech giant debuted its first model under Wang's leadership - the Muse Spark.