Samsung's Q2 profit plummets 56%, exceeding expectations, with NVIDIA's HBM3E certification issues as the main reason

Zhitong
2025.07.07 23:36
portai
I'm PortAI, I can summarize articles.

Samsung Electronics' operating profit in the second quarter plummeted 56% year-on-year to 4.6 trillion won, far exceeding analysts' expectations. Despite revenue reaching 74 trillion won, inventory pressure and setbacks in technological breakthroughs in the AI chip sector are the main reasons. Samsung's 12-layer HBM3E chip has not received certification from NVIDIA, resulting in market share being seized by competitors SK Hynix and Micron Technology. The operating profit of Samsung's chip division in the second quarter is expected to be 2.7 trillion won, still lower than 6.5 trillion won in the same period last year

According to Zhitong Finance APP, Samsung Electronics faced a setback in its second-quarter performance. Preliminary data released by the South Korean tech giant shows that its operating profit plummeted 56% year-on-year to 4.6 trillion won (approximately 3.3 billion USD), far exceeding analysts' previous expectations of a 41% decline. This marks the first profit drop for the company since the first quarter of 2023. Although revenue for the quarter reached 74 trillion won, the cost pressure from inventory backlog became a key factor dragging down performance. The complete financial report will disclose net profit and specific data for each business segment later this month.

Behind the pressure on performance is Samsung's key breakthrough in the artificial intelligence chip sector being hindered. In the high-bandwidth memory (HBM) chip market, which drives NVIDIA's (NVDA.US) AI accelerator core components, Samsung's most advanced 12-layer HBM3E product has yet to receive certification from NVIDIA, while competitor SK Hynix has seized the market opportunity, and U.S. Micron Technology (MU.US) is also accelerating its expansion. Surveys indicate that Samsung's chip division is expected to report an operating profit of 2.7 trillion won in the second quarter, which, although an increase from 1.1 trillion won in the previous quarter, still falls short of last year's 6.5 trillion won.

In April this year, Samsung released positive signals, stating that it had delivered upgraded HBM3E samples to major customers and expected to achieve mass production in the next quarter, while planning to start production of the next-generation HBM4 chips in the second half of the year. However, the reality is not optimistic: SK Hynix has already delivered the world's first batch of 12-layer HBM4 samples to customers, with Micron following suit in June, while Samsung has delayed the certification process due to design adjustments. Although Samsung recently secured HBM orders from AMD (AMD.US) and became a supplier alongside Micron, failing to obtain early certification for HBM3E from NVIDIA, the largest customer in the AI chip sector, still puts it in a passive position in the market share competition.

Market research firm Bernstein analyzed that the certification timeline for Samsung's 12-layer HBM3E has been postponed from the originally expected second quarter to the third quarter, and its HBM market share forecast has also been adjusted downward. The firm expects that by 2025, SK Hynix will maintain a leading position with a 57% market share, while Samsung (27%) and Micron (16%) will catch up, making the market landscape more balanced. Samsung's chip business head, Choi Yong-hyun, admitted at the March shareholders' meeting that early setbacks in the HBM market led to a lag behind SK Hynix and promised not to repeat the same mistakes in the HBM4 field, with this next-generation memory set to be applied in NVIDIA's Rubin GPU architecture.

Analysts at Daishin Securities believe that despite uncertainties in the certification timeline, Samsung is still expected to achieve mass production of HBM4 by the third quarter of 2025. Currently, this South Korean tech giant is facing unprecedented competitive pressure, and how to achieve technological breakthroughs and market breakthroughs in the AI memory field will be key to reversing its performance decline