Meta (formerly Facebook) has recently signed a landmark infrastructure agreement with Nebius, a cloud service provider based in Amsterdam. The agreement, which could be worth up to $27 billion, aims to ensure Meta has sufficient and advanced computing power for the next five years to cope with the intensifying global AI competition.
Key Terms of the Agreement
This long-term contract consists of two parts, reflecting Meta's strong demand for computing resources:
Core Supply: Nebius will provide Meta with $1.2 billion in dedicated computing capacity. This computing power will be deployed on a large scale using NVIDIA's upcoming Vera Rubin series AI chips, and is expected to begin delivery in early 2027.
Priority Purchase Right: Meta also committed to purchasing an additional $1.5 billion in computing capacity over the next five years. This capacity was originally intended for other Nebius cloud customers, but Meta has the right to purchase any remaining computing power first.
Nebius, an emerging "neocloud" service provider, rapidly rose after separating from Yandex-related operations. This major positive news caused Nebius' stock price to surge by more than 13% in the early trading session on Monday.
Notably, NVIDIA announced an investment of $200 million in Nebius last week, further solidifying its position as a key partner of NVIDIA. For Meta, this deal is an important part of its multi-billion-dollar AI investment plan. Previously, Meta had hinted that its AI-related capital expenditures this year could reach as much as $135 billion.
This collaboration marks Meta's accelerating shift from general-purpose data centers to dedicated AI clusters, by securing the initial quotas of top-tier chips, aiming to maintain a leading edge in the speed of large model development and deployment.
