TradingKey - Amid the exponential expansion of AI computing power, memory chips have long transcended their traditional role as computer "peripherals" to become core strategic resources that determine the inference efficiency and deployment scenarios of AI models. As High Bandwidth Memory (HBM) falls into an unprecedented "black-hole" supply shortage, the global technology industry is undergoing a structural restructuring not seen in 20 years, which also presents investors with a brand-new dimension for value assessment. For investors seeking long-term growth, the investment value of the memory sector is being redefined.
Global memory market faces supply shortage.
In 2026, as AI large models advance rapidly, High Bandwidth Memory (HBM) has become a core catalyst sparking a global transformation in the tech industry. As AI models accelerate their penetration from cloud-based training to edge-side inference, memory chips have emerged as a strategic core resource determining AI computing performance and application implementation. This fundamental shift in demand structure is dragging the global tech industry into an unprecedented "supply scramble," with everyone from chipmakers to end-device manufacturers feeling the pressure of tight memory supply.
Industry anxiety is being transmitted through various channels. MSI General Manager Golden Chiang publicly stated that 2026 is the most severe challenge the company has faced since its inception. He recalled that in the past, as long as company executives personally stepped in, they could always secure a certain amount of capacity support from suppliers. Now, however, even when CEOs of tech giants visit in person, they are often met with the response: "There really is no stock." This complete reversal of supply chain bargaining power vividly reflects the severe imbalance between supply and demand in the current memory market.
The convening of the 2026 Nvidia GTC global AI developer conference has further ignited the competitive flames in the memory market. Nvidia ( NVDA) CEO Jensen Huang officially unveiled the long-rumored Feynman architecture during his keynote speech. This next-generation computing platform will feature sixth-generation High Bandwidth Memory (HBM4E) and even more advanced HBM5 technology. The clarification of this technical roadmap will undoubtedly further intensify the technological race among the three major memory manufacturers—Samsung, SK Hynix, and Micron ( MU )—in the HBM field, accelerating the shift of industry resources toward high-bandwidth storage chips.
As core players in the global storage industry, the three major manufacturers showcased their latest collaboration results with Nvidia at the conference.
Samsung Electronics unveiled its next-generation HBM4E memory chip for the first time. This product utilizes innovative Hybrid Copper Bonding (HCB) technology, enabling chip stacks of 16 layers and above while reducing thermal resistance by more than 20%, laying the foundation for further improvements in AI computing power. SK Hynix focused on demonstrating the mass production capabilities of its HBM4 products and high-bandwidth storage solutions adapted for Nvidia's Rubin GPU. Micron Technology emphasized its breakthrough in 1́γ DRAM technology and its storage product portfolio optimized for data centers.
Through these technical demonstrations, the three manufacturers are attempting to solidify their positions at the technological high ground of the AI memory sector.
However, supply pressures in the global memory market are intensifying further. The National Samsung Electronics Union (NSEU), the largest union at Samsung Electronics, passed a resolution for a general strike with over 90% in favor, planning an 18-day collective action starting in May 2026.
This labor dispute, triggered by the wage and bonus system, will directly impact Samsung's semiconductor plant in Pyeongtaek, Seoul, and is expected to affect nearly half of the facility's capacity. Since the Pyeongtaek plant handles critical supply tasks for global DRAM and HBM products, this strike could very well evolve into a supply shock for the global chip market, further driving up memory product prices.
Simultaneously, the global memory market is entering a historic wave of price increases.
Samsung Electronics completed DRAM supply price negotiations last month, with the average price of general-purpose DRAM skyrocketing by approximately 100% compared to the previous quarter; the price of some server DRAM products increased even more. SK Hynix and Micron Technology followed suit, completing first-quarter supply contract negotiations with similar increases. The pattern of collective price hikes by the three major memory manufacturers has basically taken shape.
The core driver of this price surge is undoubtedly the explosion of the AI wave. With the rapid popularization of global AI large model training and data center expansion, memory has evolved from a common electronic component into a strategic resource supporting the development of AI computing power. Data shows that the demand for DRAM from AI servers is 8 to 10 times that of traditional servers, while the AI infrastructure capital expenditure of global tech giants is expected to reach $650 billion in 2026, an approximately 80% surge from last year. This demand growth shows no obvious inflection point and is expected to continue driving the prosperity of the memory market for several years to come.
While demand remains high, the supply gap in the memory market is widening further.
The three giants—Samsung, SK Hynix, and Micron—have all concentrated more than 80% of their advanced capacity on higher-profit AI-specific memory products, significantly squeezing the supply capacity for general-purpose DRAM. Combined with the long expansion cycles and high technical barriers inherent in memory chip production, the market will be unable to release new capacity quickly in the short term.
Alleviating fears of memory shortages
To alleviate collective market fears of shortages, Samsung Co-CEO Jun Young-hyun dropped a bombshell at a shareholders' meeting, stating that the company is considering shifting memory contracts from traditional quarterly cycles to long-term agreements of three to five years. This strategy of locking in capacity for the long term underscores that AI memory demand will remain robust for years to come.
On the same day, Samsung Electronics and AMD ( AMD) announced the signing of a memorandum of understanding (MOU) to expand strategic cooperation in next-generation artificial intelligence memory and computing technologies. Samsung will become the primary supplier of HBM4 High Bandwidth Memory chips for AMD's next-generation Instinct MI455X graphics processing units (GPUs) and provide advanced DRAM solutions for AMD's sixth-generation EPYC central processing units (CPUs), codenamed "Venice." Both parties also agreed to explore the possibility of foundry collaboration, where Samsung would provide foundry services for next-generation AMD products.
This marks AMD CEO Lisa Su’s first visit to South Korea since taking office in 2014. She also visited Naver’s headquarters, where the two companies announced an expansion of their cooperation in AI infrastructure. Su’s trip aims to strengthen partnerships with South Korean tech firms to counter competition from Nvidia in the AI chip market.
According to data from Counterpoint, Samsung holds approximately 22% of the global HBM market, while market leader SK Hynix holds a dominant 57% share. Samsung’s deep collaboration with AMD will help narrow the gap with its competitors in the fast-growing HBM sector, while also highlighting the intense competition among global chipmakers to secure advanced memory supplies.
Competitive Landscape of the Top Three Memory Giants
The global memory market is currently dominated by three industry giants: Micron Technology of the U.S., and Samsung Electronics and SK Hynix of South Korea. These three companies control the vast majority of global memory chip capacity, and their product strategies and capacity adjustments have a decisive impact on market prices and technological trends.
Leveraging its first-mover advantage in HBM technology, SK Hynix commands a 57% share of the global HBM market, making it the undisputed industry leader. The company is deeply integrated with core AI computing clients such as NVIDIA, AMD, and Google ( GOOGL ), Microsoft ( MSFT ), and other core AI computing clients, securing two-thirds of NVIDIA's Rubin-generation HBM4 orders and exclusive supply rights for Microsoft's Maia 200 chip.
The company's performance is set for explosive growth in 2026, with the average selling price of DRAM projected to surge 243% year-over-year. Operating margins are expected to soar above 70%, while return on equity (ROE) is forecast to hit a record high of over 80%.
As the world's largest memory manufacturer, Samsung Electronics is accelerating its efforts to close the technical gap in the HBM sector. In addition to debuting its next-generation HBM4E memory chips at the GTC 2026 conference, the company has signed a strategic agreement with AMD to become the primary HBM4 supplier for AMD's next-generation Instinct MI455X GPUs and will provide advanced memory solutions for AMD's sixth-generation EPYC CPUs.
Furthermore, Samsung is considering transitioning memory contracts from quarterly signings to three-to-five-year long-term agreements to secure future high-growth AI memory demand. Although Samsung's current HBM market share is only 22%, its massive production capacity and strong capital resources position it to gradually narrow the gap with SK Hynix after 2027.
Micron Technology has adopted a strategy of differentiated competition, decisively exiting the consumer storage brand business to focus all resources on high-value-added sectors like data centers and AI. This strategic shift has yielded significant results, with the company's stock price surging 168% in 2026, making it the biggest dark horse in the global memory market.
Micron has achieved a breakthrough in 1γ DRAM technology, with its memory chip performance increasing by 15% and power consumption decreasing by 20%, while its yield ramp-up speed has broken industry records. In the HBM sector, Micron's HBM capacity for fiscal year 2026 is already fully sold out, and the company plans to increase its HBM market share to over 20% by 2027.
Long-term investment value of memory stocks becomes increasingly evident.
From an investment perspective, the memory industry has shifted from traditional cyclical stocks to a new 'Davis Double Play' trajectory.
On one hand, the continuous rise in memory prices has directly driven a significant increase in manufacturer profitability; on the other hand, memory's status as a core strategic resource for AI computing power is increasingly prominent, leading market valuations to shift from cyclical stocks to tech growth stocks, with P/E ratios steadily rising.
Major investment banks have been raising earnings forecasts for the three major memory giants. Citi ( C) predicts that Samsung Electronics' operating profit will reach 155 trillion KRW in 2026, a 253% increase from the previous year; Morgan Stanley ( MS) forecasts SK Hynix's operating profit to be 225 trillion KRW in 2027; in its recently announced second quarter of fiscal year 2026, Micron Technology achieved a gross margin of 74.4%.
Although the memory market is currently at high levels, the industry supercycle is expected to last until 2027 or beyond, given the long-term growth in AI computing demand and the technical barriers to capacity expansion.
Find out more
Comments