Specifications
Product Family
DDN Infinia platform with AI2000, AI2200 and AI400X3 series appliances; EXAScaler parallel filesystem; IntelliFlash and SFA NVMe platforms
Platform Scope
Software-defined data intelligence for the full AI factory lifecycle — training, inference serving and agentic AI
AI2000 Appliance
1U all-QLC-flash system with 12 hot-swap 2.5-inch U.2 NVMe drive bays
AI2000 Networking
Up to 4x 200G network ports per node
AI2200 Appliance
Next-generation Infinia appliance delivering 2x the capacity and 2x the read performance of the prior generation for AI inference, RAG and data-intensive workloads
AI400X3 Series
AI400X3M and AI400X3 systems alongside SFA400X3 and ES400X3 platforms
SFA NVMe Platforms
SFA200NVX and SFA400NVX deliver 100% NVMe storage with scalable flash performance
Scaling Model
Scales out by adding servers to meet workload and performance demand rather than by scaling a single controller
EXAScaler
Parallel filesystem providing high-performance data access to many of the world's top AI systems and supercomputers, also available in cloud deployments
Data Management
Native metadata intelligence with extreme low-latency access, unifying structured and unstructured AI datasets
Multi-Tenancy
Multi-tenancy at scale with containerised services and security controls for multi-user AI environments
GPU-Initiated IO
DDN is demonstrating GPU-initiated data access with NVIDIA SCADA architecture on the Infinia platform, removing the CPU from the data path to cut GPU idle time
Cloud Availability
Initial availability in Amazon S3 object storage environments with Google Cloud support, expanding to more architectures over time
Deployment Context
Built for data centres, private and public clouds and the edge, serving more than 11,000 customers
Overview
DDN Infinia is a software-defined data intelligence platform built for the whole AI factory lifecycle — training, inference serving and agentic AI — rather than for a single workload tier. It is aimed at the problem that dominates large AI storage estates: data is created across many locations and sensors, must be moved to where compute is, and needs metadata intelligence attached to it so that the right bytes reach the right GPU without manual data engineering. Infinia's answer is native metadata intelligence with extremely low-latency access and a unified view over structured and unstructured AI datasets, wrapped in multi-tenancy and containerised services.
On the hardware side, DDN pairs the platform with purpose-built appliances. The AI2000 is a 1U all-QLC-flash system with twelve hot-swap 2.5-inch U.2 NVMe bays and up to four 200G network ports, designed for energy efficiency and high density rather than raw capacity per rack unit. The AI2200 is the next-generation Infinia appliance and doubles both capacity and read performance for AI inference, RAG and data-intensive workloads. Alongside them sit the AI400X3 series, the SFA NVMe platforms offering 100% NVMe storage, EXAScaler for parallel filesystem access at supercomputer scale, and the IntelliFlash portfolio. Infinia scales out by adding servers rather than by growing a controller pair, which is the model large inference estates need.
The most consequential development for AI infrastructure is GPU-initiated IO. DDN is working with NVIDIA's SCADA architecture on the Infinia platform so that GPUs request data directly rather than asking the CPU to fetch it — eliminating the idle compute time and wasted spend that CPU-mediated storage access creates in data-intensive inference. QS Compute supplies DDN AI storage platforms and appliances to data centre, HPC and AI programmes; contact us for sizing, configuration and volume pricing.
Key Benefits
One platform for the whole AI lifecycle. Training, inference serving and agentic AI on a single software-defined data layer. Density and efficiency, not just capacity. A 1U all-QLC appliance with twelve U.2 NVMe bays and up to 4x 200G. Metadata intelligence built in. Native low-latency metadata so the right bytes reach the right GPU. Ready for GPU-initiated IO. Under development with NVIDIA SCADA to take the CPU out of the data path.
Applications
AI training clusters and inference serving fleets, RAG and agentic AI pipelines requiring unified access to structured and unstructured data, high-performance computing and supercomputing storage, media and autonomous-driving data estates, research institution and government AI programmes, and enterprise data centres consolidating AI workloads onto a single scalable storage platform with multi-tenancy.
Request a Quote — DDN INFINIA — AI FACTORY STORAGE PLATFORM AND AI2000 / AI2200 ALL-FLASH APPLIANCES
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →