SK Hynix and NVIDIA are reportedly co-developing a specialized AI SSD under a project named Storage Next. This inference-optimized storage solution, targeting a 2027 release, aims to address the high-throughput, low-latency demands of running AI models by functioning as a pseudo-memory layer for massive parameters.
The planned AI SSD could reach a staggering 100 million IOPS, far exceeding current enterprise drives. This performance jump would come from new NAND and controller architectures designed for energy efficiency. A prototype from SK hynix is expected next year.
Such a product would intensify strain on NAND supply chains, already pressured by demand from cloud and AI firms. This could trigger a supply crunch similar to the current DRAM market, disrupting pricing and availability as the industry pivots to accommodate new AI infrastructure needs.
The planned AI SSD could reach a staggering 100 million IOPS, far exceeding current enterprise drives. This performance jump would come from new NAND and controller architectures designed for energy efficiency. A prototype from SK hynix is expected next year.
Such a product would intensify strain on NAND supply chains, already pressured by demand from cloud and AI firms. This could trigger a supply crunch similar to the current DRAM market, disrupting pricing and availability as the industry pivots to accommodate new AI infrastructure needs.
