从“韩国最没希望的企业”到市值8万亿 全球第二大IPO来了

This article tells the turnaround story of South Korean memory chip maker SK hynix, which grew from a bankrupt-level penny stock to a trillion-dollar market cap company and pulled off the world's second-largest initial public offering (IPO). Key takeaways are as follows:

1. Core IPO data: SK hynix listed on the Nasdaq on July 10, raising $26.5 billion — the largest IPO ever by a foreign company in the U.S., and the second-largest IPO globally in history. It closed up 12.76% on its first trading day, pushing its market capitalization to around 8.3 trillion yuan. The offering was 7x oversubscribed, with subscriptions from top global institutions totaling nearly $200 billion, reflecting strong investor confidence in the AI memory segment.

2. Key lessons from SK hynix's turnaround: The company foresaw industry demand and began developing High Bandwidth Memory (HBM) technology as early as the early 2000s. Even when the industry downturn drove massive losses, it refused to cut HBM capacity and sustained 15 years of R&D investment. It ultimately caught the wave of the AI large model boom to become the world's largest HBM supplier with a 58% market share. Its operating profit margin hit 72% in the first quarter of 2026, outperforming NVIDIA and TSMC, and the company distributed huge performance bonuses to all employees. This success demonstrates that long-term technological investment ultimately pays off when industry trends take off.

This article outlines development opportunities and pathways for brands in the semiconductor and related sectors in the AI era, with key insights as follows:

1. Product R&D and consumer trends: The boom in AI large models has transformed HBM from a niche technology into an irreplaceable core infrastructure for AI computing, driving explosive industry demand. Brands need to anticipate industry trends in advance and commit to long-term technological布局 to build hard-to-replicate first-mover advantages. SK hynix's decision to retain HBM capacity during the industry downturn, which ultimately made it a core supplier to NVIDIA, is a prime example.

2. Pathways to enhancing brand value: To address the valuation discount that plagues South Korean companies, SK hynix chose to list in the U.S. and won recognition from global capital, effectively boosting its global valuation and influence. This offers a valuable reference for outbound brands looking to enhance their corporate value.

3. Industry trends: Memory chips have been redefined from cyclical assets to core assets for AI computing. The global industry's total revenue is projected to reach $1.76 trillion by 2027, and extended supply shortages will sustain a high-growth outlook for the segment for years to come. Brands should提前布局 AI memory-related businesses.

This article summarizes the current landscape, opportunities and risks of the memory and AI-related segments for industry practitioners, with key takeaways as follows:

1. Demand shifts and growth opportunities: The growing demand for memory bandwidth driven by AI large model training has pushed the global memory chip industry into a super growth cycle. Total industry revenue is projected to reach nearly $992 billion in 2026, and double to $1.76 trillion by 2027. Industry supply shortages will extend from 2027 through 2030 and beyond, with high-value products including HBM, DRAM and eSSD seeing sustained strong demand, making this a clear high-growth segment.

2. Risk reminders: The memory industry is highly cyclical with sharp price fluctuations, and downturns can lead to massive losses. Only practitioners who提前布局 core technologies and can weather industry troughs will capture the upside of the upcycle. The segment is also highly concentrated, with the top three global players holding the vast majority of market share, so new entrants need to identify clear differentiation opportunities.

3. Key takeaways: Practitioners should maintain investment in core high-potential businesses even during industry downturns, and avoid cutting long-term R&D projects hastily. This allows them to seize first-mover advantage and earn excess returns when the market opportunity arrives.

This article outlines development directions, opportunities and insights for memory chip and related manufacturing facilities, with key takeaways as follows:

1. Product manufacturing and design requirements: In the AI era, High Bandwidth Memory (HBM) has become a core necessity for AI computing systems. HBM relies on Through-Silicon Via (TSV) 3D stacking technology. Relevant manufacturing facilities need to adjust their R&D directions to align with AI application demand,提前布局 the required cutting-edge manufacturing technologies, and adapt to shifting market demand.

2. Business opportunities: The global memory chip industry will face sustained supply shortages from 2026 to 2030, and industry revenue will double within two years, creating ample market space. Chinese memory players have risen rapidly: ChangXin Memory Technologies (CXMT) has risen to fourth place globally in DRAM market share, is seeing explosive growth, and has launched an A-share IPO. Domestic supporting factories can pursue cooperation with leading Chinese memory players to capture growth opportunities.

3. Insights for transformation and development: The memory industry faces large cyclical fluctuations. Companies should not arbitrarily cut all unprofitable projects during industry downturns; instead, they need to identify core technology projects aligned with future trends and sustain investment. This allows them to build competitive barriers and earn excess returns when the opportunity arrives, and companies should dare to布局 niche cutting-edge directions even when they do not appear mainstream.

This article outlines industry trends, opportunities and development directions for semiconductor-related service providers, with key insights as follows:

1. Industry development trends: AI has driven memory chips into a brand-new super cycle. Memory chips have transformed from traditional highly cyclical products into core assets of AI computing infrastructure. The global industry size will double to $1.76 trillion by 2027, and sustained high growth will drive rapid increases in demand for semiconductor R&D, manufacturing, finance and other related services.

2. Core new technology directions: The TSV 3D stacking technology required for HBM is the core cutting-edge technology in the current memory sector. Top global vendors are accelerating the iteration of next-generation HBM technology, driving rapid growth in demand for R&D and manufacturing supporting services related to this technology. Service providers can布局 relevant businesses in advance.

3. Customer pain points and solution directions: Memory chip makers require large amounts of long-term capital for R&D and capacity expansion, and many face valuation discounts and have demand for expanded global financing channels. Financial and capital connection service providers can develop targeted solutions to meet this need. Leading vendors are rapidly expanding capacity for core products, creating strong demand for supply chain supporting and technology R&D services. Service providers can adjust their business directions to match the needs of these leading vendors.

This article offers valuable insights for capital market platforms and semiconductor industry platforms, with key content as follows:

1. Market demand: The AI memory segment is currently one of the most sought-after sectors by global capital. SK hynix's IPO was 7x oversubscribed, with subscriptions from top global institutions totaling nearly $200 billion, demonstrating that high-quality high-growth AI semiconductor projects have strong demand for cross-border listing and financing platforms. Platforms can optimize their targeted services to attract relevant high-quality projects to list.

2. Platform development directions: Nasdaq holds strong attraction for high-quality global technology projects, and many high-quality companies choose to list in the U.S. to eliminate valuation discounts and obtain higher valuations. Platforms can learn from this experience and optimize their service systems for technology companies to improve their attractiveness to high-quality projects.

3. Risk mitigation and new opportunities: Memory chips are a highly cyclical industry, and even amid the current high growth there is still risk of future cyclical fluctuations. Platforms need to fully disclose these risks when bringing in relevant projects. At the same time, Chinese memory vendors are rising rapidly, and a large number of Chinese memory companies have demand for listing and financing. Platforms can connect and布局 these opportunities in advance to capture the new growth driven by domestic substitution.

This article provides the latest industry developments and research materials for semiconductor industry researchers in the AI era, with key insights as follows:

1. New industry developments: The boom in AI large models has redefined the industrial attributes of memory chips. What was once a highly cyclical industry is now a core infrastructure for AI computing, entering a brand-new super growth cycle. SK hynix currently leads the global HBM market with a 58% share, while Chinese player ChangXin Memory Technologies has risen to fourth place in global DRAM market share, with explosive profit growth: it has transformed from a persistent loss-maker to nearly 300 million yuan in daily profit. The global semiconductor memory industry landscape is undergoing rapid restructuring.

2. New industry characteristics: The global memory chip industry is about to enter a multi-year period of sustained supply shortages. SK hynix's CEO projects 2027 will be the most supply-constrained year for the industry, with shortages extending beyond 2030. Industry size will double within two years, which marks a complete departure from the traditional cyclical pattern of frequent downturns, and creates new research topics.

3. Research insights: SK hynix sustained 15 years of R&D investment in a niche cutting-edge technology, and continued investment even during the industry trough, ultimately earning excess returns. This breaks the traditional business model where memory companies profit purely from cyclical fluctuations. Additionally, the decision of a leading South Korean technology company to list in the U.S. to escape the "Korea discount" provides an entirely new typical case for research on valuation differences across global capital markets and corporate listing location choices.