Today, we are proud to announce that Cloudian HyperStore 8.2.6 has achieved the Foundation level of NVIDIA-Certified Storage designation—giving enterprises a verified, high-performance object storage solution purpose-built for AI workloads at scale.
The era of the AI factory has arrived. Data centers are no longer passive warehouses for storing bits and bytes—they are active manufacturing facilities for intelligence. AI models are being trained, fine-tuned, and served at unprecedented scale, and the infrastructure feeding them must evolve to match.
In this new paradigm, data is the fuel and storage is the fuel pump of the AI Factory. If the pump cannot keep pace with the engine, performance stalls—no matter how powerful the GPUs. That is why the validation of storage infrastructure has become mission-critical for enterprises making the transition from AI experimentation to AI production.
What Is NVIDIA-Certified Storage—and Why Does It Matter?
NVIDIA-Certified™ Storage empowers AI factories with high-performance, reliable data solutions. This rigorous certification ensures storage systems meet the highest standards of quality, reliability, and the strict performance required for training and inference, and advanced AI pipelines like RAG and agentic AI. Customers can confidently build AI factories that leverage data for faster, more accurate, and reliable AI models, unlocking new possibilities and driving innovation across industries. By achieving the Foundation level of NVIDIA-Certified Storage, Cloudian HyperStore 8.2.6 has been validated to meet essential demands of AI workload involving up to 128 GPUs.
Object Storage for AI: Fast, Scalable, and Simple
Cloudian HyperStore was designed to meet exactly the demands that AI workloads place on storage infrastructure. The NVIDIA certification validates what HyperStore customers already know: it delivers the three core attributes enterprise AI requires.
Fast: Terabytes per Second, When It Matters Most
HyperStore is engineered for extreme throughput. A single rack can scale to terabytes per second of sustained bandwidth—enough to keep even the most demanding GPU clusters fully fed during training runs. Whether processing petabytes of unstructured data for model training or serving thousands of concurrent inference requests, HyperStore eliminates storage as the performance bottleneck. The NVIDIA certification validates this performance across the precise I/O patterns AI workloads demand: sequential reads, random I/O, and low-latency retrieval.
Scalable: From Terabytes to Exabytes, Non-Disruptively
AI data grows fast—and HyperStore grows with it. Its modular, distributed architecture scales non-disruptively from terabytes to exabytes, across one or multiple sites, managed as a single unified system. There are no forklift upgrades, no migration headaches, and no scheduled downtime. As your AI ambitions scale, your storage scales alongside them—seamlessly and continuously.
Simple: One Pool for All Your Data
Data fragmentation is one of the biggest hidden costs in enterprise AI. Datasets scattered across cloud buckets, NAS systems, legacy archives, and siloed applications represent untapped fuel for AI models—but only if you can access them efficiently. HyperStore’s universal S3-native API and workload consolidation capabilities let you bring data from multiple storage types and sources into a single, scalable pool. The result: reduced cost, eliminated complexity, and more of your data available to AI pipelines—right when you need it.
Bridging Cloud and On-Premises: S3 Compatibility and Data Repatriation
One of HyperStore’s most strategic advantages for AI teams is its ability to bridge the public cloud and on-premises worlds. With industry-leading S3 API compatibility, HyperStore integrates natively with virtually every cloud-native AI framework, data pipeline tool, and S3-enabled application—no code changes required, no re-engineering of existing workflows.
This matters enormously for enterprises seeking to repatriate data from the public cloud. Many AI workloads began in the cloud—where data already resided—but as model training and inference scale up, the economics shift decisively toward on-premises deployment. Data sovereignty requirements further accelerate this shift, as regulated industries and government customers increasingly demand that sensitive data remain under their direct control.
Cloudian’s integrated data migration technologies make the transition from cloud to on-premises straightforward. Enterprises can move datasets back on-premises, unlocking previously inaccessible data for AI inference, fine-tuning, and advanced analytics—while retaining the S3 interfaces their teams and tools already depend on. The result: the cost efficiency and data sovereignty of on-premises AI, without sacrificing the developer experience or ecosystem compatibility of the cloud.
A Comprehensive, Certified AI Infrastructure Stack
The NVIDIA-Certified Storage designation builds on and deepens Cloudian’s growing collaboration with NVIDIA. Cloudian recently launched the HyperScale® AI Data Platform (AIDP), built on the NVIDIA AI Data Platform reference design, which pairs HyperStore’s S3-native object storage with NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs to deliver a fully integrated, on-premises AI platform—enabling enterprises to unlock the 80% of data that was previously inaccessible to AI.
Together, the NVIDIA-Certified Storage Foundation designation and the HyperScale AIDP give customers a comprehensive, validated AI infrastructure stack—from high-performance data storage to GPU-accelerated intelligence. For enterprises building production-grade AI Factories, this means fewer unknowns, faster deployment, and greater confidence in the infrastructure underpinning their AI strategy.
The Bottom Line
As AI moves from experimentation to production, the storage layer can no longer be an afterthought. The AI factory demands infrastructure that is:
- Fast enough to feed accelerated compute at rack-scale terabytes-per-second throughput.
- Scalable enough to grow non-disruptively from today’s deployments to tomorrow’s exabyte-scale ambitions.
- Simple enough to consolidate data from across the enterprise into a single, manageable pool.
- Compatible enough to migrate data and workloads from the public cloud with ease.
Cloudian HyperStore delivers all of the above—and now, with the NVIDIA-Certified Storage designation, delivers proof. Enterprises have a verified, production-ready path to AI Factory storage that is certified for the most demanding GPU-accelerated environments in the world.
