Nebul Launches Five Specialized Storage Platforms
The cloud provider is diversifying its storage backend to better handle distinct AI training and inference requirements.
Updated on Oct. 26, 2026 in Data Centers
Live Poll
Do you believe using specialized storage infrastructure improves performance for your business applications?
Cloud provider Nebul has formalized its storage architecture by offering five distinct platforms to support GPU-intensive workloads. This announcement outlines how the company matches specific storage technologies to varied performance requirements.
Why it matters
By diversifying its storage backend, Nebul moves away from one-size-fits-all infrastructure to better manage distinct AI workloads like high-throughput training and inference. This strategy allows the cloud provider to tailor performance to specific operational needs.
Nebul now offers five storage platforms to support AI GPU clients, including DDN Infinia for KV cache and DDN EXAScaler for large-scale training. Integration with NVIDIA technologies, such as Dynamo and NIXL, remains a core component of this architectural approach.
The players
Nebul
A European cloud provider that specializes in GPU-accelerated computing infrastructure and AI-native storage architectures.
NVIDIA
A leader in GPU design and AI computing hardware whose proprietary technologies, such as Dynamo and NIXL, are integrated into Nebul storage stacks.
The details
Nebul selects storage architectures based on workload requirements rather than relying on a single underlying system. DDN Infinia serves as an AI-native data lake for inference and key-value (KV) cache workloads—temporary storage that holds prompt responses to speed up repeated queries. For high-throughput requirements during large-scale model training, it employs DDN EXAScaler, while NetApp ONTAP and Novus provide enterprise file and parallel file systems, respectively. VAST Data rounds out the offerings with an integrated data platform that combines database and data engine capabilities.
Timeline
October 26, 2026: Nebul clarified its five storage platform offerings.
The Tech Race
Nebul's strategy highlights the escalating competition to optimize data throughput for high-demand AI training and inference. This move marks a departure from the industry trend toward monolithic storage, favoring instead a disaggregated approach that aligns specific platforms with unique GPU requirements.
GPU-using clients can now match their specific inference or training workflows to specialized storage backends, potentially improving performance metrics for complex AI tasks. This multi-platform support is currently available, though Nebul notes it may rotate or add new storage providers for specific tasks like KV caching in the future.
The takeaway
Nebul is shifting the bottleneck of GPU-cloud operations away from monolithic storage to specialized, high-performance engines. Users should watch for future technical benchmarks comparing these five platforms to see which configurations offer the best cost-to-performance ratio for their training cycles.
Further reading
Explore deeper into the evolving architecture of Data Centers powering global artificial intelligence infrastructure.
Source note: This article includes information reported by Blocksandfiles.
Live Poll
Do you believe using specialized storage infrastructure improves performance for your business applications?
