Browser not supported. Please use Google Chrome or Microsoft Edge to access Ringle.
#Tech
Trends in the Semiconductor Industry
The Crisis of IDMs
Updated: 2024.11.05
4 min read · Advanced
material_image
Trends in the Semiconductor Industry

The advent of AI has driven profound shifts in the semiconductor industry. South Korea’s Samsung Electronics, renowned for its substantial [1] revenue from DRAM sales, and the U.S.-based Intel, once the leader in CPU sales, both reported disappointing results this year. Analysts note that both companies have struggled to keep pace with rapidly evolving semiconductor market trends. Until 2022, these two companies held the top global positions in semiconductor sales. But how has AI reshaped the dynamics of this industry?

The company that surpassed Samsung and Intel in semiconductor sales was Taiwan’s TSMC. Integrated Device Manufacturers (IDMs) like Samsung and Intel manage the entire semiconductor process—from design to manufacturing and post-processing. In contrast, TSMC operates as a pure-play foundry, focusing solely on manufacturing for fabless companies specializing in semiconductor design. A notable example of a fabless firm is NVIDIA, whose primary manufacturer is TSMC.

AI has played a pivotal [2] role in increasing the prominence of foundries. AI data centers require AI accelerators that rely on specific semiconductors: HBM and GPU. HBM, composed of vertically stacked DRAM chips, is critical to determining the performance of AI accelerators. GPUs, once secondary to CPUs, have gained importance due to their efficiency in handling large-scale data processing, essential in the AI era.

NVIDIA dominates the AI accelerator market, capturing about 90% of the market share. While Samsung focused on general-purpose DRAM products, NVIDIA sought out SK Hynix for customized HBM manufacturing. Meanwhile, the GPUs used in AI accelerators are primarily produced by TSMC, while Intel has struggled with stagnation [3] in CPU advancements. Because of the advanced technology, foundries that can customize for each customer started to gain traction.

Although both IDMs also operate their foundries, they lag behind [4] TSMC, which focuses exclusively on manufacturing. Because IDMs must invest across the entire production process, they are often limited in R&D investment, hindering their capacity for innovation. Recently, Intel's self-produced chips underperformed compared to TSMC’s offerings, even jeopardizing Intel’s competitiveness in CPUs. Consequently, Intel has outsourced the production of its next-generation CPU, “Lunar Lake,” to TSMC.

Prior to the 2000s, most global semiconductor players were IDMs. However, as semiconductor technology diversified over the last two decades, IDMs have faced increasing challenges, especially as competitors sharpen their focus on specialized areas. Critics also argue that the bureaucratic [5] structures of large corporations have become obstacles to fast and innovative decision-making. So, how will IDMs like Samsung and Intel strategize to overcome this crisis and adapt to the shifting landscape?

반도체 산업의 동향

AI의 등장은 반도체 산업에 큰 변화를 가져왔습니다. DRAM 판매로 막대한 수익을 올리는 것으로 유명한 한국의 삼성전자와 한때 CPU 판매의 선두주자였던 미국의 인텔은 모두 올해 실망스러운 실적을 발표했습니다. 분석가들은 두 회사 모두 빠르게 진화하는 반도체 시장 트렌드를 따라잡는 데 어려움을 겪고 있다고 지적합니다. 2022년까지만 해도 이 두 회사는 반도체 판매량에서 세계 최고의 자리를 차지했습니다. 하지만 AI는 이 업계의 역학 관계를 어떻게 재편했을까요?

반도체 매출에서 삼성과 인텔을 뛰어넘은 기업은 대만의 TSMC였습니다. 삼성과 인텔과 같은 종합 반도체 업체 (IDM)는 설계부터 제조, 후공정까지 전 반도체 제작 과정을 관리합니다. 이에 반해 TSMC는 반도체 설계를 전문으로 하는 팹리스 기업을 위한 제조에만 집중하는 순수 파운드리로 운영됩니다. 팹리스 기업의 대표적인 예로는 NVIDIA가 있으며, TSMC가 이들의 주요 제조업체입니다.

AI는 파운드리의 위상을 높이는 데 중추적인 역할을 했습니다. AI 데이터 센터에는 특정 반도체에 의존하는 AI 가속기가 필요합니다: 바로 HBM과 GPU입니다. 수직으로 적층된 DRAM 칩으로 구성된 HBM은 AI 가속기의 성능을 결정하는 데 매우 중요합니다. 한때 CPU의 부차적인 반도체로 여겨졌던 GPU는 AI 시대에 필수적인 대규모 데이터 처리의 효율성으로 인해 그 중요성이 커지고 있습니다.

엔비디아는 AI 가속기 시장에서 약 90%의 시장 점유율을 차지하며 독보적인 위치를 차지하고 있습니다. 삼성이 범용 DRAM 제품에 집중하는 동안, 엔비디아는 맞춤형 HBM 제조를 위해 SK하이닉스를 찾았습니다. 인텔이 CPU 발전의 정체로 어려움을 겪고 있는 반면, AI 가속기에 사용되는 GPU는 주로 TSMC가 생산하고 있습니다. 고도화된 기술로 인해 고객사별 맞춤 제작이 가능한 파운드리가 각광을 받기 시작한 것입니다.

두 IDM 모두 파운드리도 운영하고 있지만, 제조에만 집중하는 TSMC에 비해 뒤처져 있습니다. IDM은 전체 생산 공정에 걸쳐 투자해야 하기 때문에 R&D 투자에 제한을 받아 혁신을 발휘하지 못하는 경우가 많습니다. 최근에는 인텔이 자체 생산한 칩이 TSMC의 제품에 비해 성능이 떨어지면서 인텔의 CPU 경쟁력까지 위협받고 있습니다. 그 결과 인텔은 차세대 CPU인 '루나 레이크'의 생산을 TSMC에 아웃소싱했습니다.

2000년대 이전에는 대부분의 글로벌 반도체 업체들이 IDM이었습니다. 그러나 지난 20년 동안 반도체 기술이 다양해지면서, 특히 경쟁업체들이 전문 분야에 집중하면서 IDM은 점점 더 많은 도전에 직면하고 있습니다. 또한 비평가들은 대기업의 관료적 구조가 빠르고 혁신적인 의사결정을 방해하는 장애물이 되었다고 주장합니다. 삼성과 인텔과 같은 IDM은 이러한 위기를 극복하고 변화하는 환경에 적응하기 위해 어떤 전략을 세울까요?

Discussion Questions
Q1
In your own words, please briefly summarize the article.
여러분의 언어로 교재를 간단히 요약해 주세요.
Q2
What part of the reading resonated with you most?
이번 교재에서 가장 공감하는 내용은 무엇인가요?
Q3
How has the rise of AI changed the dynamics of the semiconductor industry, particularly for IDMs like Samsung and Intel?
AI의 부상은 반도체 산업, 특히 삼성과 인텔과 같은 IDM의 역학 관계를 어떻게 변화시켰나요?
Q4
What are the advantages and disadvantages of the IDM model (design-to-manufacturing integration) compared to the pure-play foundry model?
순수 파운드리 모델과 비교했을 때 IDM 모델(설계-제조 통합)의 장점과 단점은 무엇인가요?
Q5
Do you think focusing on a specific area could benefit IDMs in the long run?
특정 분야에 집중하는 것이 장기적으로 IDM에 도움이 될 수 있다고 생각하시나요?
Q6
How might the bureaucratic structures of large corporations limit their ability to innovate in a fast-evolving industry like semiconductors?
대기업의 관료적 구조가 반도체처럼 빠르게 진화하는 산업에서 혁신 능력을 어떻게 제한할 수 있을까요?
Q7
Considering the challenges IDMs face, what strategies do you think Samsung and Intel could adopt to regain competitiveness in the industry?
IDM이 직면한 어려움을 고려할 때, 삼성과 인텔이 업계에서 경쟁력을 회복하기 위해 어떤 전략을 채택할 수 있다고 생각하시나요?
Q8
If you have a question or questions that you'd like to discuss during your class, please write them down.
궁금한 점이 있거나 수업 중에 얘기해 보고 싶은 질문이 있으면 적어주세요.
Expressions
substantial
of considerable size, importance, or worth
Example
1

The company made a substantial investment in new technology.

Example
2

Her contributions to the project were substantial and led to its success.

pivotal
of crucial importance in relation to the development or success of something
Example
1

The discovery was pivotal to advancing our understanding of genetics.

Example
2

He played a pivotal role in negotiating the peace agreement.

stagnation
a state of no growth, movement, or development
Example
1

Economic stagnation over the last few years has affected the job market.

Example
2

The company feared stagnation if it didn't innovate its products.

lag behind
to fall behind in progress, development, or achievement compared to others
Example
1

Our company is lagging behind competitors in terms of technology.

Example
2

She felt she was lagging behind her classmates in mathematics.

bureaucratic
related to the complex rules, procedures, and rigid structures typical of administrative systems
Example
1

The process of getting a permit was slow and bureaucratic.

Example
2

He was frustrated with the bureaucratic system that delayed the project.

본 교재는 당사 편집진이 제작하는 링글의 자산으로 저작권법에 의해 보호됩니다. 링글 플랫폼 외에서 자료를 활용하시는 경우 당사와 사전 협의가 필요합니다.

The advent of AI has driven profound shifts in the semiconductor industry. South Korea’s Samsung Electronics, renowned for its substantial [1] revenue from DRAM sales, and the U.S.-based Intel, once the leader in CPU sales, both reported disappointing results this year. Analysts note that both companies have struggled to keep pace with rapidly evolving semiconductor market trends. Until 2022, these two companies held the top global positions in semiconductor sales. But how has AI reshaped the dynamics of this industry?

The company that surpassed Samsung and Intel in semiconductor sales was Taiwan’s TSMC. Integrated Device Manufacturers (IDMs) like Samsung and Intel manage the entire semiconductor process—from design to manufacturing and post-processing. In contrast, TSMC operates as a pure-play foundry, focusing solely on manufacturing for fabless companies specializing in semiconductor design. A notable example of a fabless firm is NVIDIA, whose primary manufacturer is TSMC.

AI has played a pivotal [2] role in increasing the prominence of foundries. AI data centers require AI accelerators that rely on specific semiconductors: HBM and GPU. HBM, composed of vertically stacked DRAM chips, is critical to determining the performance of AI accelerators. GPUs, once secondary to CPUs, have gained importance due to their efficiency in handling large-scale data processing, essential in the AI era.

NVIDIA dominates the AI accelerator market, capturing about 90% of the market share. While Samsung focused on general-purpose DRAM products, NVIDIA sought out SK Hynix for customized HBM manufacturing. Meanwhile, the GPUs used in AI accelerators are primarily produced by TSMC, while Intel has struggled with stagnation [3] in CPU advancements. Because of the advanced technology, foundries that can customize for each customer started to gain traction.

Although both IDMs also operate their foundries, they lag behind [4] TSMC, which focuses exclusively on manufacturing. Because IDMs must invest across the entire production process, they are often limited in R&D investment, hindering their capacity for innovation. Recently, Intel's self-produced chips underperformed compared to TSMC’s offerings, even jeopardizing Intel’s competitiveness in CPUs. Consequently, Intel has outsourced the production of its next-generation CPU, “Lunar Lake,” to TSMC.

Prior to the 2000s, most global semiconductor players were IDMs. However, as semiconductor technology diversified over the last two decades, IDMs have faced increasing challenges, especially as competitors sharpen their focus on specialized areas. Critics also argue that the bureaucratic [5] structures of large corporations have become obstacles to fast and innovative decision-making. So, how will IDMs like Samsung and Intel strategize to overcome this crisis and adapt to the shifting landscape?

*This material is designed for the exclusive use of Ringle students on the Ringle platform.