Microsoft’s Nebius GPU Infrastructure Agreement

ZUSAMMENFASSUNG

In 30 Sekunden

Matthew Berman describes an agreement for Nebius to provide Microsoft with GPU infrastructure and connects it to demand for training and inference. He identifies Nebius as the sponsor and also promotes its AI Studio service. Nebius’s September 2025 filing says the approximately $17.4 billion contract value through 2031 was subject to deployment and availability conditions; additional services could raise it to about $19.4 billion.

1The Microsoft agreementBerman reports the five-year Nebius deal as worth $17.4 billion. The company filing makes that value conditional on deployment and availability.
2Compute for AI workloadsBerman connects the agreement with infrastructure needs for pre-training, post-training, and inference.
3A sponsor’s serviceBerman promotes Nebius AI Studio for model inference and fine-tuning; the segment is sponsored by Nebius.

Das Wichtigste

Nebius, the video’s sponsor, announced a five-year agreement to provide Microsoft dedicated GPU capacity.
The approximately $17.4 billion contract value was conditional on deployment and availability, according to the company’s 2025 filing.
Microsoft could purchase additional services or capacity, potentially raising the stated value to about $19.4 billion.
Berman links infrastructure demand to model training, post-training, and inference.
The video’s discussion of Nebius AI Studio is sponsored promotion, not an independent product evaluation.

ThemenInfrastruktur & Deployment AI cost & performance Modelle & Training

VOLLTEXT

Den bearbeiteten Text nach Kapiteln lesen

Transkript zeigen

Nur das Transkript

A GPU infrastructure agreement

Berman introduces the Microsoft–Nebius deal and identifies Nebius as the video’s sponsor. The contract value is subject to deployment and availability conditions.

FOLIE 1 / 2Nebius’s 2025 Microsoft announcementThe company announcement introduces the AI infrastructure agreement.

Microsoft has signed an agreement with Nebius, the sponsor of this video, for AI infrastructure. I describe it as a $17.4 billion, five-year deal. [Nebius’s September 2025 filing](https://assets.nebius.com/assets/86625727-9e66-46c1-af79-c27d57434e3c/25-25580-1_Nebius%2520Group%2520N.V._6-K.pdf%3Fcache-buster%3D2025-09-08T21%3A23%3A09.625Z) states that this value through 2031 was subject to GPU service deployment and availability; additional purchases could increase it to about $19.4 billion.

AI workloads and Nebius services

He discusses training and inference needs, then promotes Nebius AI Studio. The short does not independently assess the service.

Microsoft is building its own models, which require substantial computing capacity for pre-training, post-training, and inference. I describe Nebius as a source of capacity alongside Azure. The public agreement provides for dedicated GPU infrastructure, but the video does not specify which particular Microsoft models or workloads will use it.

FOLIE 2 / 2Nebius AI Studio model listingThe sponsored segment shows a historical model listing with service and pricing information.

I also promote Nebius AI Studio for people building with open models, including inference and fine-tuning. Nebius sponsors this segment, so the recommendation is promotional rather than an independent comparison of services. The video mentions a $20 credit offer from the time; its availability and terms may change.