A Samsung Electronics Co. 12-layer HBM3E, high, and different DDR modules organized in Seoul, South Korea, on Thursday, April 4, 2024.
SeongJoon Cho | Bloomberg | Getty Images
Samsung Electronics was as soon as the dominant participant in a sort of semiconductor referred to as reminiscence, placing it in an incredible place to capitalize on the growth of synthetic intelligence.
But the South Korea electronics big has now fallen behind its long-time rival SK Hynix in next-generation chips which have been key part for AI silicon chief Nvidia. The end result? Samsung’s revenue has plunged, round $126 billion has been wiped off its market worth, in response to information from S&P Capital IQ, and an govt issued a uncommon public apology in regards to the firm’s current monetary efficiency.
Memory is a essential sort of chip used to retailer information, and it may be present in a plethora of units from smartphones to laptops. For years, Samsung was the simple chief on this expertise, forward of South Korean rival SK Hynix and U.S. competitor Micron.
But as AI functions akin to OpenAI’s ChatGPT rose in reputation, the underlying infrastructure required to coach the large fashions they depend on turned a much bigger focus. Nvidia has emerged as the highest participant on this house with its graphics processing items (GPUs) which have grow to be the gold normal utilized by tech giants for AI coaching.
An important a part of that semiconductor structure is high-bandwidth reminiscence, or HBM. This subsequent technology of reminiscence includes stacking a number of dynamic random entry reminiscence (DRAM) chips, but it surely had a small market earlier than the AI growth.
That’s the place Samsung obtained caught out and failed to speculate.
“HBM has been a very niche product … for a long time and Samsung has not focused its resources on its development,” Kazunori Ito, director of fairness analysis at Morningstar, advised CNBC by electronic mail.
“Due to the difficulty of the technology involved in stacking DRAMs and the small size of the addressable market, it was believed that the high development costs were not justified.”
SK Hynix noticed this chance. The firm aggressively launched HBM chips which had been authorized to be used in Nvidia structure and, within the course of, the South Korean agency established an in depth relationship with the U.S. big. Nvidia’s CEO even requested the corporate to hurry up provide of its subsequent technology chip, underscoring the significance of HBM to its merchandise.
SK Hynix posted file quarterly working revenue within the September quarter.
“With strong R&D (research and development) investments and established industry partnerships, SK Hynix maintains an edge in both HBM innovation and market penetration,” Brady Wang, related director at Counterpoint Research, advised CNBC by electronic mail.
Samsung advised CNBC that, within the third quarter, whole HBM gross sales grew greater than 70% quarter-on-quarter. The tech big added that the present product referred to as HBM3E is in mass manufacturing and producing gross sales.
The South Korean tech firm famous that growth for its next-generation HBM4 is “underway according to plan” and that the corporate is concentrating on beginning “mass production” within the second half of 2025.
Can Samsung make a comeback?
Analysts mentioned that Samsung is lagging behind opponents for a variety of causes, together with underinvestment in HBM and the truth that it isn’t a first-mover.
“It is fair to say that Samsung has not been able to close the gap with SK Hynix on the HBM development roadmap,” Morningstar’s Ito mentioned.
Samsung’s skill to make a comeback within the brief time period seems to be intently linked to Nvidia.
An organization should cross a strict qualification course of earlier than Nvidia approves it as a HBM provider — and Samsung has not but accomplished this verification. But a inexperienced mild from Nvidia may open the door for Samsung to return to development and compete extra successfully with SK Hynix, in response to analysts.
“Since NVIDIA holds more than 90% of the AI chip market, where most HBMs are used, NVIDIA’s approval is critical for Samsung to benefit from the robust demand for AI servers,” Ito mentioned.
A Samsung spokesperson mentioned that the corporate has made “meaning progress” concerning HBM3E and has “completed an important phase in the qualification process.”
“We expect to start expanding sales in the fourth quarter,” the spokesperson mentioned.
Meanwhile, Wang famous that Samsung’s power in analysis and growth, in addition to the corporate’s semiconductor manufacturing capability that would assist it catch as much as SK Hynix.