Meta Platforms (NASDAQ: META) recently made a big announcement that caused neocloud stocks like Nebius (NASDAQ: NBIS) and Iren (NASDAQ: IREN) to stumble. Facebook’s parent company said it would explore building its own neocloud business and renting out its excess compute capacity.
That aligns with the company’s earlier announcement of intent to build tens of gigawatts of AI data centers this decade and hundreds of gigawatts of capacity over time. However, investors shouldn’t rush to imagine Meta Platforms running a vibrant neocloud division under its corporate umbrella.
Missed Nvidia in 2009? This Rare Signal Is Flashing Again. In 2009, a “Double Down” signal flashed for a little-known chipmaker called Nvidia. For the first time in years, that same “Total Conviction” signal is flashing for a company 1/100th the size of Nvidia.Continue »
Meta Platforms needs more compute
Companies like Nebius and Iren aren’t using the data center capacity they’re building to develop their own large language models or physical AI systems. They are AI enablers.
Meta Platforms still needs computing power for its own initiatives, such as AI models and smart glasses. It’s also using artificial intelligence to enhance the effectiveness of the advertising on its platforms to drive more revenue.
Meta Platforms signed a bunch of deals last year to obtain more computing power to support its AI ambitions, and that trend continued this year. It has already inked extended agreements with CoreWeave and Nebius.
The fact that Meta Platforms is still scrambling to sign these types of deals shows that it won’t have available compute capacity to lease to other companies anytime soon. Other hyperscalers and AI giants are also ramping up these types of deals. For instance, Anthropic recently secured 401 megawatts of capacity from Terawulf via a 20-year, $19 billion contract.
There isn’t enough cloud computing capacity in Meta Platforms’ grasp to run its own AI businesses and supply more to outside enterprises. As long as this gap exists, Meta Platforms won’t have an immediate path into the neocloud industry.
AI data centers take years to build
Meta Platforms CEO Mark Zuckerberg has expressed his desire to build many AI data centers for the company, but each of those projects will take years to complete. For instance, Meta Platforms’ Hyperion project won’t be completed until 2030. That will be a massive 5-gigawatt (5,000-megawatt) data center upon completion, but considering how Meta Platforms’ demands for in-house compute are rising, it’s tough to predict how much of that capacity it will be able to make available for smaller enterprises.
Comments are closed, but trackbacks and pingbacks are open.