| 会社名 | Group | 親会社 | 子会社 | 得意先 | 仕入先 | 業務提携 | 銀行等 | 株主等 |
|---|---|---|---|---|---|---|---|---|
| SK hynix NAND Product Solutions Corp. |
SK hynix held its first Partners Day in Japan, gathering over 20 key local partners to share mid‑to‑long‑term business and technology roadmaps. The company aims to secure AI‑era memory supply, strengthen Korea‑Japan semiconductor cooperation, and enhance stable supply chains through early information sharing and closer collaboration.
AI data centers increasingly face data-transfer bottlenecks rather than limits in GPU speed. Copper links consume power and lose signal at high speeds and over longer distances, prompting a shift to optical interconnects using silicon photonics. Co-packaged optics places optical engines next to switch chips to shorten electrical connections, increase bandwidth, and reduce power use.
SK hynix provided a link to access its Q3 2026 earnings conference call. Email verification is required.
SK hynix reported record third-quarter 2026 results, with revenue exceeding 100 trillion won, driven by strong demand for AI memory and HBM4. The company expects memory shortages to continue through 2030 and plans to expand advanced DRAM and NAND capacity. Long-term agreements cover more than half of its 2027–28 output, and the company plans about 100 trillion won in shareholder returns, including gains from its Kioxia investment.
Solidigm has stopped reviewing a potential IPO amid weak market and share-price conditions. It will continue exploring alternative financing to fund U.S. fab capacity for growing AI memory demand and support expansion and SK hynix shareholder value.
AI performance increasingly depends on the entire data center, not just individual chips. Future infrastructure is being designed through co-design across compute, memory, storage, networking, power, and cooling, from servers to racks and full facilities. Shared standards and early collaboration, particularly around memory such as HBM, are becoming important.
AI performance is shifting from raw compute toward data movement and placement, as growing transformer workloads make memory bandwidth and hierarchy major cost factors. HBM, advanced packaging, and PIM bring compute closer to data, making memory-centric, system-level design a new basis for AI semiconductor competitiveness.
SK hynix Inc. will announce its third quarter 2026 earnings via a press release before 9:00 a.m. on October 28, 2026 (Seoul time). A Korean‑English interpreted conference call will follow at 9:00 a.m.
SK hynix presented its next-generation AI infrastructure vision at OCP Global Summit 2026, highlighting memory-centric server systems. The article notes that the explosive growth in AI data has created data bottlenecks as a new challenge.
SK hynix developed 3D FeNAND-based analog compute-in-memory technology that combines data storage and computation to reduce AI data movement. By co-optimizing cells, interconnects and page schemes, it achieved 10–20 times higher throughput and 4,000–5,000 times better energy efficiency than conventional 3D NAND.