NGX Rolls Out ExaScale Block Storage for AI Workloads

Why It Matters

ExaScale gives enterprises a high‑performance, cost‑effective storage option for AI workloads, challenging entrenched U.S. vendors and accelerating Turkey’s emergence in the enterprise‑storage market.

Key Takeaways

• ExaScale scales from terabytes to exabytes for AI workloads
• Uses NVMe, RDMA, and three‑way replication for low latency
• Supports on‑premises and public‑cloud deployments
• Competes with Dell, HPE, NetApp, offering price advantage
• Targets enterprises in Turkey, Middle East, and Europe

Pulse Analysis

The explosion of artificial‑intelligence models has forced organizations to rethink data‑storage strategies. Traditional SANs and file‑based solutions struggle to keep pace with the petabyte‑scale datasets required for training and inference. NGX’s ExaScale array addresses this gap by offering a pure block storage platform that can expand to exabyte levels, delivering the throughput and latency needed for modern AI pipelines while preserving a unified management layer for mixed workloads.

Technically, ExaScale adopts a disaggregated architecture reminiscent of VAST Data, separating high‑write‑speed SSDs for ingest and metadata from a capacity tier built on QLC SSDs. NVMe over Fabrics, including RDMA and TCP, provides ultra‑fast east‑west traffic, while three‑way replication and RAID‑6 protect data without a quorum node. This kernel‑bypass, user‑space design reduces overhead, enabling sub‑millisecond response times that are critical for real‑time analytics and model serving.

From a market perspective, NGX’s channel‑only sales model and Turkish origin give it a pricing edge against Dell, HPE, NetApp and other U.S. competitors. By delivering enterprise‑grade performance at lower cost, ExaScale can attract cost‑sensitive customers in emerging markets while still appealing to global enterprises seeking diversified supply chains. As DRAM and NAND prices rise, NGX’s decision to limit price hikes may further solidify its position, making it a compelling option for organizations looking to scale AI workloads without inflating capital expenditures.