DDN Shatters Performance Records with AI400X3, Setting New Standard in AI Infrastructure at MLPerf Storage 2025

DDN, the global leader in AI and data intelligence solutions, has unveiled the AI400X3 storage appliance, setting a new industry standard for performance and efficiency in AI infrastructure. The system has achieved record-breaking results in the latest MLPerf Storage v2.0 benchmarks, validating its ability to deliver exceptional performance density in a compact, energy-efficient 2U form factor. Powered by DDN’s EXAScaler parallel file system, the AI400X3 is designed to accelerate the most demanding AI workloads at scale. It excels in both single-node and multi-node configurations, simulating real-world deployment scenarios from initial AI projects to large-scale distributed training environments. In both cases, the appliance delivered outstanding performance, effectively saturating hundreds of simulated H100 GPUs across diverse AI workloads. The MLPerf Storage 2025 benchmark is widely regarded as the gold standard for measuring how well a storage system can support AI training. The AI400X3’s results demonstrate that it can keep pace with—and even outperform—the most advanced GPUs, all within a small, power-efficient footprint. This capability removes a key bottleneck in AI development: data starvation. In single-node testing, the AI400X3 delivered top-tier throughput, enabling rapid model training with minimal infrastructure overhead. For organizations scaling to hundreds of GPUs, the multi-node results showcase consistent, high-speed data delivery under extreme load—ensuring GPUs remain fully utilized throughout training cycles. This leads to faster time-to-insight, improved training efficiency, and reduced operational costs. The appliance’s compact design and low power consumption make it ideal for data centers facing tight constraints on space, power, and cooling. By delivering high performance in a small footprint, the AI400X3 supports sustainable, scalable AI growth. Sven Oehme, CTO at DDN, emphasized the significance of the results: “AI at scale demands more than raw power—it requires precision-engineered infrastructure. The AI400X3 proves that DDN can deliver relentless performance, efficiency, and reliability, all while helping organizations train AI models faster and more cost-effectively.” DDN has a long-standing reputation as a trusted partner in high-performance computing and AI. Since 2016, NVIDIA has relied exclusively on DDN to power its internal AI clusters, a testament to the company’s proven ability to deliver mission-critical infrastructure. The AI400X3 is already enabling breakthroughs across industries—from decoding the human genome and advancing medical imaging to training complex vision models. By eliminating data bottlenecks and ensuring continuous GPU utilization, DDN’s solution transforms raw data into real-time intelligence. With validation from MLPerf, DDN provides enterprises and AI innovators with the confidence to build faster, train smarter, and deploy at scale. The AI400X3 isn’t just storage—it’s purpose-built infrastructure that drives the next era of AI innovation. For more information, visit ddn.com.