A CNBC report reveals that Samsung Electronics, once the global leader in memory chips, now faces a challenging landscape as rival SK Hynix has advanced with high-bandwidth memory (HBM) technology essential to artificial intelligence (AI).
As AI-driven applications like ChatGPT have spiked demand for high-performance chips, SK Hynix has emerged as the go-to supplier for Nvidia, the dominant player in AI GPUs.
With a sharp decline in profits and a significant drop in market valuation, Samsung must confront strategic gaps in next-gen memory development to stay competitive in the AI landscape.
Samsung’s missed opportunity in high-bandwidth memory
Memory chips are central to various devices, from smartphones to data centres, storing and processing vast amounts of information.
For years, Samsung led the industry in memory technology, overshadowing both SK Hynix and the US-based Micron.
Yet, as demand for high-bandwidth memory (HBM) surged, driven by the rise of AI, Samsung’s lack of investment in this area became increasingly evident.
High-bandwidth memory involves stacking multiple dynamic random access memory (DRAM) chips to enhance speed and capacity, ideal for powering sophisticated AI models.
Nvidia, seeking to meet the demands of tech firms using AI, turned to SK Hynix, which had anticipated this trend.
In response, SK Hynix aggressively accelerated HBM production, receiving Nvidia’s approval for use in its GPUs, a move that strengthened its market position.
The decision not only solidified SK Hynix’s lead in HBM but also positioned it as a critical supplier to Nvidia, further widening the gap with Samsung.
As a result, S&P Capital IQ data states that Samsung has experienced a dramatic fall in its profits, losing around $126 billion in market value, while noting that Samsung’s late entry into HBM has been a costly oversight.
Samsung’s bid to catch up in HBM innovation
Despite Samsung’s struggles, the tech giant has shifted its strategy to close the gap.
In the third quarter, Samsung’s HBM sales rose by 70% compared to the previous quarter, driven by strong demand in AI sectors.
Samsung’s HBM3E chip has reached mass production, and its next-generation HBM4 is slated for launch by late 2025, indicating a renewed commitment to competing in this vital sector.
Yet, it is best to remain cautious about Samsung’s ability to regain its former dominance.
With SK Hynix holding a significant advantage in the HBM space and having established deep industry partnerships, Samsung’s path forward remains challenging.
Samsung’s future success in HBM hinges on accelerated research and development (R&D) efforts and the timely delivery of its next-gen memory technology.
The question remains whether these moves can allow Samsung to reclaim its foothold in the rapidly evolving AI market.
The post Nvidia HBM delay deepens Samsung’s $126 billion AI problem appeared first on Invezz