Illustrating a shift from HPC capability to AI capacity
This marks a milestone as DDN makes its data engines, EXAScaler and Infinia, available software-defined
Introducing new Inference Economics Scorecard to measure inference infrastructure by the useful work it delivers, as applied in an eight-hour benchmark with Nebius on Nvidia HGX B300 GPUs and Weka NeuralMesh
Advancements address critical data movement, processing and storage challenges for AI and HPC infrastructure
Kingston demonstrates memory and storage product support for data center
CPU-based NVMe RAID arrives on AMD EPYC platforms in February 2027, with no dedicated RAID controller required
New operations services offering provides single point of accountability that reduces operational risk and delivers sustained productive GPU capacity
Vera delivers higher performance for agentic science and 2x greater energy efficiency than x86 CPUs
New open dataset and post-training recipes help researchers adapt Nemotron 3.5 Super VL for life sciences and chemistry
Highlighting power and cost efficiency, based on 8-LPU servers
Up to 7.3x faster on the shared-file and metadata work that keeps GPUs waiting: no new hardware, no proprietary client, open-source Ceph underneath
The next frontier of AI infrastructure is moving beyond raw compute performance toward high-speed interconnects and seamless coordination across computing nodes
Brings more than a decade of supercomputer-scale networking and consecutive first-silicon successes to mainstream AI; new Ethernet SuperNIC adds low-latency RoCEv2 to standard 800G switches, with Ultra Ethernet to follow