• 구독신청
  • My스크랩
  • 지면신문
FNTIMES 대한민국 최고 금융 경제지
ad

6th Generation HBM Competition... Samsung Electronics 'All-in-One' vs SK Hynix 'Super Gap'

김재훈 기자

rlqm93@fntimes.com

기사입력 : 2024-08-14 15:35 최종수정 : 2024-08-14 15:56

Samsung Electronics and SK Hynix Declare Production of 6th Generation HBMs Next Year!
Samsung Electronics aims to 'turn the tide' with foundry technology specializing in fine processes.
SK Hynix Strengthens HBM Super Gap Strategy with D-RAM Expansion and Major Investments.

Samsung Electronics' Seocho office building (left) and Hynix's Icheon headquarters (right) / Photo from each company

Samsung Electronics' Seocho office building (left) and Hynix's Icheon headquarters (right) / Photo from each company

[Korea Financial Times, Kim JaeHun] The competition between Samsung Electronics and SK Hynix to develop next-generation AI semiconductor memory HBM4 (High Bandwidth Memory, 6th Generation) products, which is scheduled to be released next year, is gradually taking shape. Samsung Electronics, which has lost pride in SK Hynix in its 5th generation HBM3E product, is aiming for a "big jump" by banking on its all-in-one strength such as fine process foundry. SK Hynix, which currently has the largest global share of HBM, plans not to allow a reversal by taking the reins of super-gap power.

According to Samsung Electronics and SK Hynix on 13th August, the two companies will start mass production of HBM4, the next-generation HBM product, next year. HBM is categorized into generations based on the level of stacking technology (stacking) of memory. Up to now, up to the fifth generation has been commercialized.

5th generation HBM3E is taking the lead by supplying about 90% to Nvidia, which monopolizes AI semiconductors, after SK Hynix succeeded in commercializing it for the first time in the world in March. According to market research firm Trendforce, SK Hynix is expected to take the top spot in the global share of HBM in terms of shipments this year with about 53%.

2nd place (42% share) Samsung is currently undergoing HBM3E performance testing from Nvidia and is expected to begin mass production later this year.

Samsung Electronics, which has lost its pride in SK Hynix, which it considered to be one step below it in the memory market, plans to turn the tables in the competition for HBM4 leadership. As Nvidia, which has a 98 percent monopoly in the global AI semiconductor market, releases new AI GPU 'Rubin' in 2026, it plans to quickly secure supply. Nvidia plans to install HBM4 in Rubin for the first time.

In particular, industry observers believe that Samsung will have an advantage because HBM4 requires a finer process and specific components and designs than previous products. Up to the HBM3E, the HBM was arranged horizontally around the GPU but, HBM4 is characterized by its increased computational speed by vertically stacking HBMs on top of GPUs. In this case, the 'logic die', an essential part that connects the GPU and HBM, requires foundry (semiconductor contract manufacturing) capabilities.

Samsung Electronics is the only all-in-one company in the global semiconductor market with its own memory, foundry, and packaging capabilities. This means that Samsung Electronics has strengths in cost reduction, productivity, and customized strategies compared to SK Hynix, which outsources production to Taiwanese foundry company TSMC.

"We are developing HBM4 with the goal of launching it in the second half of next year," said Kim Jae-joon, head of strategic marketing(Vice President) at Samsung Electronics' memory division, in a conference call on the 31st of last month. "We are discussing detailed specifications with our customers, such as developing customized products to meet customer-specific HBM demand, and will respond in a timely manner based on our supply capabilities."

SK Group Chairman Choi Tae-won who visited SK Hynix's Icheon headquarters on the 5th Aug. / Photo by SK Hynix

SK Group Chairman Choi Tae-won who visited SK Hynix's Icheon headquarters on the 5th Aug. / Photo by SK Hynix


On the other hand, SK Hynix plans to strengthen its technology super gap strategy, such as expanding investment and early commercialization, based on its market share.

SK Hynix is known to have placed an order for equipment to expand the M16 plant in Icheon, which is under construction as a state-of-the-art DRAM plant. The plant produces general-purpose DRAMs used in various AI and electronic devices, including next-generation HBMs.

Prior to this, SK Group announced its plan to invest about 103 trillion won in strengthening HBM competitiveness with SK Hynix over the next five years by securing an additional 80 trillion won by 2026 along with the AI-centered rebalancing declaration in June.

SK Group Chairman Choi Tae-won also met with Nvidia CEO Jensen Huang in April and TSMC Chairman Wei Zhejia in June to solidify the three-way alliance. He also encouraged members by visiting the Icheon Campus, SK Hynix's headquarters, on the 5th of this month.

Chairman Choi Tae-won said, "SK Hynix is now recognized in the HBM market, but we should not settle for the present as competition will become fiercer when 6th generation HBM is commercialized next year. We should focus on securing technological competitiveness without shaking and think fiercely about next-generation products." "Let's commercialize 6th generation HBM early next year to protect Korea's AI semiconductor leadership and contribute to the national economy."

Kim JaeHun (rlqm93@fntimes.com)

데일리 금융경제뉴스 FNTIMES - 저작권법에 의거 상업적 목적의 무단 전재, 복사, 배포 금지
Copyright ⓒ 한국금융신문 & FNTIMES.com

가장 핫한 경제 소식! 한국금융신문의 ‘추천뉴스’를 받아보세요~

산업 다른 기사

1 LG CNS, 삼송 데이터센터 '액체냉각' 인프라 구축…AI 팩토리 기반 마련 LG CNS가 네이버클라우드와 함께 고밀도 GPU 시대에 대응하기 위한 차세대 데이터센터 냉각 인프라 구축에 나선다.26일 LG CNS에 따르면, 양사는 기존 공랭 방식만으로 감당하기 어려운 고성능 GPU의 발열 문제에 대응하기 위해 삼송 데이터센터에 DTC 방식 액체냉각 인프라를 구축한다.LG CNS는 데이터센터 설계·구축·운영(DBO) 역량과 AI 인프라 기술을 결합해 DTC 액체냉각, GPU 서버 운영 환경, 전력 공급 체계, AI 플랫폼 운영 기반까지 통합 구현한다.LG CNS는 삼송 데이터센터의 위탁운영을 맡고 있으며, 네이버클라우드와 삼송 데이터센터 코로케이션 계약도 체결했다. 이번 협력을 통해 LG CNS는 삼송 데이터센터에 액체냉각 기술 중 2 CJ올리브네트웍스 유인상 대표, AX로 체질 개선…연임 기대감↑ 유인상 CJ올리브네트웍스 대표가 취임 후 3년간 사업 포트폴리오를 재편하고 AX 중심 성장 기반을 마련했다. 그의 연임 가능성에 관심이 쏠리고 있다.유 대표는 대외 사업 확대와 AX 사업을 통해 그룹 계열사 중심이던 사업 구조를 다변화하고 체질 개선을 이끌었다는 평가다.글로벌 ICT 전문가 유인상 대표…AI·빅데이터·클라우드 주력유 대표는 LG CNS 출신 글로벌 정보통신기술(ICT) 전문가다. LG CNS에서 디지털 플랫폼(IoT·데이터허브), 디지털 스페이스(스마트빌딩·스마트팜), 스마트 SOC(스마트시티·스마트교통) 등 디지털 시티·모빌리티 분야 사업을 두루 경험했다.그는 지난 2023년 7월 CJ올리브네트웍스 대표이사로 선임됐다. 유 3 ‘신차 100종 투입’ 현대차 “2030년 글로벌 555만대 판매‧영업이익률 9% 달성” 현대자동차가 지정학적 위협, 글로벌 경쟁 심화에도 공격적인 판매 및 재무 목표를 제시했다. 특히 2030년까지 100종의 신차를 투입해 글로벌 판매 555만 대, 영업이익률 9% 회복을 달성한다는 구상이다.이를 위해 글로벌 연간 생산 능력을 확대하고 북미, 유럽 등 지역별 맞춤 차량과 파워트레인 경쟁력을 제고해 완성차 본연의 기업가치를 끌어 올린다는 방침이다.2030년까지 공격적인 신차 출시, 권역별 성장전략 추진현대차는 26일 서울 여의도동 콘래드서울호텔에서 투자자와 애널리스트, 신용평가사 담당자 등을 대상으로 ‘2026 CEO 인베스터 데이’를 개최하고, 중장기 전략과 재무 계획을 발표했다.이날 행사에는 호세 무뇨스 대표이사(
ad
ad
ad

한국금융 포럼 사이버관

더보기

FT카드뉴스

더보기
[그래픽 뉴스] ISA 대개편! 나에게 유리한 계좌는?
[그래픽 뉴스] 미국 증시 새로운 키워드 'MANGOS'
환전·로또·육아휴직까지 하반기부터 달라지는 제도 TOP11
[그래픽 뉴스] 은퇴후 30년 부모님 세대의 생존전략
[그래픽 뉴스] 퇴근 후 주차했는데 수익 발생? V2G의 정체

FT도서

더보기