Newsletter
Nebius Group shares soared over 60% in extended trade after the Nvidia GPU provider for AI training inked a deal with Microsoft worth up to $19.4 billion over five years.
Nebius Group runs data centers that provide powerful Nvidia graphics chips, giving companies the computing power they need to train and run AI models.
The deets: Nebius, headquartered in Amsterdam, provides cloud infrastructure powered by Nvidia GPUs, which are the core chips used for training large AI models like ChatGPT.
Under the deal, it will supply Microsoft with this high-performance computing power from its New Jersey data center. The agreement spans till 2031 with a base value of $17.4 billion, and Microsoft has the option to scale it up by another $2 billion.
This partnership directly helps Microsoft meet the exploding demand for AI compute capacity across Azure and OpenAI workloads.
The why: demand for AI infrastructure is outstripping supply, with OpenAI and other cloud clients scrambling for GPU capacity. Microsoft has already tapped CoreWeave and Google to plug the gaps. For Nebius, the deal means faster growth and financing options to scale GPU services in multiple tranches this year and next.
Big theme: Computing resources are basically the muscles of modern tech, made up of processors (CPUs and GPUs), memory, storage, and networking that together provide the power to run apps, crunch data, and train AI models.
The global AI compute infrastructure market is exploding. Demand for GPUs and cloud computing resources far outstrips supply, pushing companies like Microsoft, Google, and Amazon to partner with specialized players such as CoreWeave and Nebius.
By 2030, analysts project the broader AI infrastructure market could cross $300–400 billion, driven by cloud, data centers, and GPU capacity.
