AIN Family

Overview

The AIN (AI NAND) Family is SK hynix's line of AI-optimized NAND storage products, comprising three distinct products mapped to the performance-capacity-cost triangle of AI inference requirements. AIN P (Ultra High Performance SSD) targets ultra-high IOPS fine-grain random access for vector database search, delivering 50M IOPS at 512-byte access on PCIe Gen6 — 7x more than a conventional Gen6 enterprise SSD (7M IOPS at 4KB). AIN B (High Bandwidth Flash) complements HBM by stacking up to 16 NAND cores on a base/logic die with an interposer, delivering HBM-like bandwidth with SSD-like capacity to address LLM inference decode and KV cache demands. AIN D (High Capacity/Low Cost SSD) is a JEDEC NL-SSD-compliant product targeting NL-HDD replacement while addressing data center power and front-space constraints.

AIN B's key differentiator is enabling inference batch size and token length to scale simultaneously without a hard tradeoff, allowing larger LLM weights to run on a single GPU node without scale-out. It also improves Time to First Token (TTFT) in multi-turn inference by retaining the growing KV cache in-place without spilling to CPU memory or conventional SSD. SK hynix's roadmap targets AIN P E3.x PoC samples by end-2026 and 100M IOPS production capability by end-2027, with a future unified architecture integrating AIN P, AIN B, xPU, and HBM.