Samsung Electronics has reported a 56% year-on-year drop in operating profit for the second quarter of 2025, marking its first decline over a year.
The South Korean technology giant announced an operating profit of 4.6 trillion won , or about $3.3 billion, which was short of market expectations. Analysts polled by Bloomberg had expected a less severe drop of roughly 41%. Revenue for the quarter was 74 trillion won.
Several factors contributed to the earnings slump. The tech firm blamed one-time inventory costs and delays in completing shipments of its most advanced memory chips. These problems weighed on the performance of its semiconductor division, a crucial profit center for the company.
The company’s chip-making business has also been struggling for much of the past year, but the company said that operating losses are expected to be lower in the second half of the year. The increase is also anchored to a longer-term, gradual rebounding of global demand for semiconductors, including those for AI and high-performance computing.
The firm intends to publish a comprehensive full earnings report later this month, with all net income figures accompanied by detailed breakdowns by business division.
Samsung falls behind in AI chip race
Samsung’s hardest task at the moment is its underperforming business in the thriving artificial intelligence chip market. The company has been playing catch-up with its biggest domestic competitor, SK Hynix , which has a big lead over Samsung in producing high-bandwidth memory (HBM) chips.
These are very important chips for AI hardware systems, especially for companies like Nvidia, which control so much of the AI training and inference chips market. Samsung had expected to win major contracts from Nvidia for its latest chip — the 12-layer HBM3E — but hasn’t yet received final quality certification.
By contrast, SK Hynix announced a few days ago that it has begun the early shipping of 1.8V 12-Layer HBM4 samples to clients. This coup has elevated SK Hynix to Nvidia’s supplier of choice, giving it a massive head start in the competitive AI memory space. US vendor Micron Technology has also made inroads, sending out HBM samples in June.
Analysts say that postponing that date could have long-term repercussions. The Samsung note from Bernstein and analyst Mark Li cut Samsung’s projected HBM market share. Now, they expect SK Hynix to have approximately 57% of the HBM market in 2025, while Samsung has 27% and Micron 16%.
Samsung prepares to rebound in second half
Still, Samsung is hopeful. The company started shipping better HBM3E samples, nicknamed Superman, out to key customers, including Nvidia, hoping that production would finally commence if the final stages of certification were passed. The tech firm also said it plans to start mass production of HBM4 chips, the next generation of high-bandwidth memory, in the second half of 2025.
At its annual shareholders’ meeting in March, Samsung’s president of semiconductor business, Jun Young-hyun, admitted that the company failed to secure an early lead in the HBM market. He promised this would not happen with HBM4, which will feature prominently in Nvidia’s next-gen Rubin GPU architecture.
Although certification delays remain a concern, analysts have expressed optimism that Samsung could catch up. In a June report, Daishin Securities analyst Ryu Hyung-keun said that while Samsung’s 12-layer HBM3E certification process has taken longer than expected, the company appears to be on track for a strong HBM4 rollout in the third quarter of 2025.
Samsung is betting large on this upturn. The company’s long-term strategy is to lessen dependence on consumer electronics and smartphones for sales and place greater emphasis on advanced memory chips, AI infrastructure, and foundry services. If it can overcome its challenges now and execute its HBM4 roadmap, it could be competitive again in the cutthroat contest for leadership in the AI chip market.
But for now, SK Hynix is unquestionably in the lead — a position Samsung is waging a fierce battle to take back.
