Image source: Shutterstock

Highlights

  • WEKA and Nebius have partnered to deliver a high-performance GPU-as-a-Service (GPUaaS) solution for AI training and inference.

  • The integrated platform enables scalable compute and storage with ultra-low latency and high throughput, ideal for modern AI workloads.

  • A top research institution has deployed the solution, leveraging a multi-thousand-GPU cluster with 2PB of WEKA storage for advanced AI model development.

WEKA, a leader in AI-native data platform technology, has joined forces with Nebius (NASDAQ:NBIS), a rapidly emerging AI infrastructure provider, to deliver a powerful GPU-as-a-Service (GPUaaS) solution designed for next-generation artificial intelligence workloads. The collaboration brings together WEKA’s high-performance, software-defined data storage platform and Nebius’ full-stack AI cloud infrastructure.

The integrated solution is tailored for organizations conducting large-scale AI training and inference tasks, which often grapple with limitations in compute, memory, and data management. With growing demand for  specialized AI infrastructure, many enterprises and research institutions are moving toward neocloud environments.

Nebius AI Cloud stands out in this emerging category by providing a cost-optimized, AI-focused cloud environment. The platform is built to support workloads across sectors — from startups to large enterprises and research institutions. To further enhance the performance and scalability of its premium offering, Nebius selected WEKA’s data storage software. The deployment provides rapid access to data at scale, facilitating operations from petabytes to exabytes.

One of the early adopters of the new solution is a prominent research institution engaged in large-scale experimentation and AI model development. The institution selected Nebius' AI Cloud to support its intensive workloads and reserved a multi-thousand-GPU cluster optimized for AI/ML training. To meet its unique performance and management needs, the organization requested the integration of WEKA’s platform, having experienced previous success with its features such as user-level and directory-level quota management.

As part of the implementation, 2 petabytes (PB) of WEKA storage were deployed alongside Nebius’ compute resources.