{
  "id": 9076951,
  "title": "SK hynix Targets AI Inference Bottlenecks With New Memory Technologies",
  "url": "https://urgent.news/2026/09/22/sk-hynix-targets-ai-inference-bottlenecks-with-new-memory-technologies",
  "topic": "tech",
  "section": "Tech",
  "published": "2026-09-22T04:02:40.000Z",
  "source": {
    "name": "Korea IT Times",
    "slug": "korea-it-times",
    "url": "https://www.koreaittimes.com/news/articleView.html?idxno=157443"
  },
  "original_language": "en",
  "account": "SK hynix is broadening its AI memory approach beyond high-bandwidth memory (HBM) as the surge in AI inference and agent-based services generates new demands for data storage and transfer. The company is focusing on technologies such as High Bandwidth Flash (HBF), Processing-In-Memory (PIM), and SALT-KV to improve AI system efficiency. Rather than solely enhancing HBM performance, this strategy integrates various memory types and storage solutions. SK hynix displayed these technologies at the AI Infra Summit 2026 in Santa Clara, California, which attracted around 6,000 attendees. The company also doubled the size of its exhibition space from the prior year. While GPUs and HBM have driven the AI semiconductor hype, the rise of commercial AI services poses a new challenge. As user interaction with AI models increases, the data retention and retrieval requirements grow significantly. Filling this gap with HBM can be costly and impractical due to capacity constraints. SK hynix is addressing this issue through through-silicon via (TSV) technology, combining higher storage capacity with higher bandwidth. This new memory tier bridges the gap between HBM and traditional solid-state storage, making it increasingly relevant for inference workloads where large data volumes must remain readily accessible without requiring HBM-level performance. Additionally, SK hynix is advancing PIM, which integrates computing functions directly into memory, reducing latency and power consumption in AI workloads. The company also presented SALT-KV, a software-based approach that distributes previously processed data across HBM, DRAM, and SSDs based on factors like reuse likelihood and storage cost. By demonstrating these technologies at the summit, SK hynix aims to expand its role beyond HBM, focusing on how these memory products are deployed within AI systems. CEO Kwak Noh-jung emphasized SK hynix's intention to create new possibilities for AI through next-generation memory, building on its established position in HBM. SK hynix's strategic investments, including its semiconductor cluster in Yongin, South Korea, and an advanced packaging production and R&D base in Indiana, USA, underscore the company's commitment to addressing the evolving memory needs of AI systems.",
  "summary": "SK hynix is extending its artificial intelligence memory strategy beyond high-bandwidth memory as the rapid growth of AI inference and agent-based services creates new demands for storing and moving data.The company has recently put greater emphasis on technologies including High Bandwidth Flash (HB",
  "key_points": [
    "SK hynix targets AI inference bottlenecks with new memory technologies",
    "Focuses on HBF, PIM, and SALT-KV to improve AI system efficiency",
    "Demonstrates these technologies at AI Infra Summit 2026 in California"
  ],
  "editors_take": "SK hynix's expansion into new memory technologies such as High Bandwidth Flash, Processing-In-Memory, and SALT-KV positions the company to better address AI inference bottlenecks and gain a broader role in AI systems.",
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}