Lightbits Labs, best known for inventing NVMe over TCP and being a front runner in modern, software-defined cloud data platforms, announced that Nebul selected the company to provide storage for its advanced AI cloud services.

Nebul's AI cloud services, aimed at unifying data, deploying NVIDIA (News - Alert)-based Private AI and extracting actionable insights, serves companies across the EU.

With Lightbits, Nebul has the ability to deliver AI cloud services that perform 16 times better while reducing costs compared to other solutions.

By combining high performance and low latency, as well as compatibility with common orchestration environments such as Kubernetes, VMware, OpenShift, and OpenStack, Lightbits provides an optimal choice for cloud builders and service providers who are managing diverse, performance-sensitive workloads at scale.

"With Lightbits in place, Nebul can now provide EU-based organizations with a specialty AI cloud that runs on NVIDIA AI Enterprise with certified tools, frameworks, and AI apps to meet the demands of today’s most intensive applications,” said Eran Kirzner, CEO and co-founder of Lightbits. “The data platform is incredibly flexible running in container environments, like Kubernetes and Azure Kubernetes Solution (AKS), delivering accelerated performance and efficiency for cloud-native applications at scale. We attribute this versatility combined with the unparalleled speed, scalability, and cost-efficiency to the increase in interest and cloud service use cases, like AI.”

Nebul identified the need for high-performance block storage for latency-sensitive AI workloads, such as RAG model training and inference. These workloads, built on vector, real-time, and other NoSQL databases, demand exceptional performance at scale.

The Lightbits cloud data platform scales beyond the petabyte level and delivers up to 75 million IOPS with consistent sub-millisecond tail latency, even under heavy load - making it ideal for managing real-time AI application data and storing training parameters and tags.