South Korea’s SK Group, the parent company of Nvidia‘s chip supplier SK Hynix, has announced plans to amass 80 trillion won ($56 billion) by 2026. The funds will be primarily directed towards investments in artificial intelligence (AI) and semiconductors.
What Happened: The company’s top brass agreed during a weekend meeting that the funds, secured through improved profitability, business structure optimization, and synergy creation, would be invested in high-bandwidth memory (HBM) chips, data centers, and personalized AI assistant services. The strategy is designed to bolster its supply chains for emerging technology, Nikkei Asia reported on Monday.
The group’s chipmaking unit, SK Hynix, is projected to control over half the global market for HBM this year. The unit supplies nearly all HBM chips used by Nvidia NVDA, which utilizes them to boost the efficiency of its graphics processing units for AI computing and training.
However, Samsung Electronics and Micron are also vying for a significant share in the HBM market, posing potential competition for SK Hynix.
Despite the looming competition, analysts opine that SK Hynix continues to lead in chip-stacking technology, a technique that amalgamates multiple conventional DRAM chips to enhance speed and facilitate large-scale data processing.
SK Group has yet to disclose detailed plans on how it aims to reach its 80-trillion-won target.
Why It Matters: In April, SK Hynix reported its first profitable quarter in five, attributing the turnaround to soaring demand for AI chips. The company’s strong performance was driven by increased sales of AI server products, particularly high-bandwidth memory, a technology crucial for AI chipsets used by Nvidia.
However, in June, Nvidia was in the final stages of certifying Samsung’s HBM chips, a critical step before Samsung could supply components crucial for training AI platforms. Nvidia’s approval is essential for companies like Samsung and Micron to compete with SK Hynix in the AI chip market.
Image via SK Hynix
This story was generated using Benzinga Neuro and edited by Pooja Rajkumari