AI Storage Solutions Under Pressure to Perform and Save Power
AI storage solutions are specialized systems that combine high-throughput media, smart data services, and GPU-friendly architectures to move and protect large training and inference datasets while keeping energy consumption, heat, and physical footprint under control across data centers, studios, and edge locations. At Computex, storage vendors are responding to a sharp rise in AI workloads by rethinking how NAS and shared storage are designed. Promise Technology, Asustor, QNAP, and Infortrend all highlight different parts of the puzzle: lowering power draw in GPU-heavy environments, shifting to all-flash NAS storage for content creation, adding on-premises AI functions to NAS operating systems, and building edge to cloud infrastructure that keeps compute close to data. Together they point to a clear trend: high-performance AI no longer excuses wasteful hardware designs, and sustainable data centers now depend on more efficient storage stacks.
Promise Technology Targets Greener AI Datacenters
Promise Technology is putting sustainability at the center of its new AI storage portfolio, with platforms tuned for high performance and lower energy use. The BoosTrak series combines dual compute and storage nodes with all-NVMe media and 100/200/400GbE plus RDMA/RoCE to keep latency low and GPUs fed without sprawling server footprints. The VTrak 8206 NVMe all-flash system adds PCIe Gen5 and NVMe-oF support for AI-driven applications and real-time analytics, again favoring dense performance over raw scale-out. For video-heavy AI workloads, the Vess A8340 integrates multi-GPU compute, Intel Xeon 6 processors, and storage in one chassis for intelligent surveillance and analytics. On the creative side, Pegasus5 Thunderbolt 5 systems push up to 6,000 MB/s for 4K and 8K workflows. GreenBoost 2.0 underscores the efficiency push by cutting storage power usage so GPUs stay busy instead of power-hungry.
Asustor Makes All-Flash NAS Storage the Default for Creators
Asustor is repositioning NAS from background file server to primary engine for content creation by standardizing on all-flash NAS storage in its Flashstor family. The design is all-flash, zero noise, and built to remove network and disk bottlenecks that slow modern studios. For music production, selected Flashstor models add S/PDIF optical output and HDMI ports, sending digital audio directly to playback equipment without extra converters. The Flashstor 6 targets advanced enthusiasts with Intel processing and dual 2.5GbE ports that can be combined via SMB Multichannel to double transmission performance. The Flashstor 12 Pro scales up to 12 M.2 slots and 10GbE networking so large music or media libraries stream almost instantly. For heavier 3D and video workloads, the Flashstor 6 Gen2 introduces 10GbE and USB4 as standard, aligning creative performance needs with more efficient, compact all-flash designs.
QNAP Adds On-Premises AI and Stronger Protection to QuTS hero
QNAP’s QuTS hero h6.0 beta focuses on reliability, data integrity, and on-premises AI features as part of its ZFS-based NAS OS. Dual-NAS High Availability now covers more systems, letting administrators pair two NAS units into an active-passive cluster that maintains services if one fails. High Availability Manager also allows clusters to connect to JBOD enclosures, so capacity can grow without sacrificing redundancy. Immutable snapshots lock snapshot data for a fixed period, blocking modification or deletion and improving ransomware readiness. QNAP integrates KMIP-based centralized key management so NAS devices can use enterprise key servers that align with FIPS 140-3 requirements. Additional tools such as FIDO2 passkeys, Secure Boot, and the announced Ransomware Guard and Secure IP Access further reduce operational risk. These capabilities help make on-premises AI deployments more secure and resilient without defaulting to energy-hungry, overprovisioned hardware.
Infortrend Builds Edge to Cloud Infrastructure for Sustainable AI Scale
Infortrend is focusing on end-to-end AI infrastructure that stretches from edge devices to private cloud data centers. Its edge computing platform brings AI inference, real-time analytics, and automation closer to where data is generated, cutting the need to send everything back to central sites and reducing network and compute overhead. It can ship as bare-metal hardware or a pre-integrated system with Standalone Edge, HA Edge, and Advanced Edge cluster modes, matching different resilience and scale needs. In private cloud environments, the enterprise cloud platform combines high-performance compute, GPU acceleration, and a pre-installed software stack to simplify deployment of AI training, big data analytics, and HPC. Storage-heavy AI and HPC tasks are served by EonStor GS 5000U, which delivers 125GB/s throughput and 2.4M IOPS, and the GSx 5000 parallel file solution, which aggregates throughput from up to 10 appliances for concurrent dataset access.
