The rapid expansion of artificial intelligence (AI) in recent years has placed unprecedented pressure on the semiconductor industry, particularly on the market for AI memory chips. As demand surges, the world’s largest technology companies are rushing to secure production capacity years in advance, while manufacturers respond with record-breaking investments. Despite expectations of eventual stabilization, high energy costs, intensifying Chinese competition, and the risk of oversupply continue to cloud the outlook.
The AI investment boom has not only propelled the shares of companies like Nvidia but has also fundamentally reshaped the memory chip market. South Korean giants SK Hynix and Samsung Electronics assert that the rapid expansion of AI data centres has created a sustained demand unlike any seen in previous semiconductor cycles. Buyers are no longer negotiating prices; they are actively seeking to lock in supply agreements that extend for years ahead, a significant shift from the traditional practices that dominated the market.
At the centre of this seismic shift are essential memory technologies such as DRAM (Dynamic Random Access Memory) and high-bandwidth memory (HBM), which are crucial for operating large language models and data centre inference systems. Without these technologies, AI systems struggle to operate efficiently. However, the supply of these components is struggling to keep pace with extraordinary demand.
SK Hynix, a critical supplier to Nvidia and the world’s second-largest memory chipmaker, characterizes the current moment as a ‘structural shift’ in the industry. Customers are pursuing long-term contracts, often spanning three to five years, to secure their supply. This marks a radical departure from the short-term agreements that previously characterized the market. The looming fears of shortages have effectively handed significant pricing power to suppliers.
Historically, the memory chip market experienced extreme volatility, oscillating between periods of oversupply leading to price collapses and instances of shortages provoking price spikes. However, this current demand surge is not driven by cyclical consumer electronics sales but rather fueled by the vast financial resources of tech giants like Microsoft, Amazon, Google, and Meta—all vying for a competitive edge in AI.
This extraordinarily high demand has given rise to what some executives call an unprecedented ‘supercycle.’ Samsung has announced ambitious investment plans, pledging around 110 trillion won to enhance production capacity and research capabilities. Similarly, SK Hynix is ramping up capital expenditures, with analysts predicting that supply constraints will persist until at least 2028. The complexity and time required for building advanced semiconductor facilities further complicate the outlook.
The concentration of the market significantly strengthens the position of producers. Samsung, SK Hynix, and Micron Technology collectively dominate approximately 90% of the global DRAM market, granting them substantial pricing power. This has led to remarkable profitability, as evidenced by SK Hynix’s operating margins soaring above 70%, while Samsung’s memory division also reports remarkably high returns.
The ramifications of this investment boom extend far beyond the technology sector. AI-centric investments are reshaping the global economy, with projections estimating infrastructure spending could soar into hundreds of billions of dollars annually. This surge stimulates various industries, ranging from construction to energy, each becoming increasingly intertwined with advancements in AI technology.
Meanwhile, semiconductor shortages have risen to the level of macroeconomic concern. Inadequate chip supply can delay data centre projects, slow down cloud development, and inflate costs for businesses—all of which could stymie productivity growth. Previous chip shortages in the automotive sector illustrate the potential for even minimal supply disruptions to cause widespread industry halts. AI’s impact could amplify such effects on a much larger scale.
Memory chips are progressively being regarded as a strategic resource, akin to oil or natural gas in prior eras. This change is reflected in the growing geopolitical tensions surrounding semiconductor supply chains, particularly between the United States and China. Policies encompassing export controls, domestic manufacturing subsidies, and total supply chain restructuring underscore the increasing importance of chip independence as nations aim to secure their technological futures.
The AI boom is also leaving its mark on energy markets. AI data centres require staggering amounts of electricity, indicating that rising chip demand is closely correlated with escalating energy consumption. Should energy prices remain high, technology companies may have to reconsider their investments, potentially triggering larger market corrections across the board.
As we look to the horizon, a pressing question arises: will AI begin to transition from data centres into everyday devices? If on-device AI proliferates, it could unleash another wave in the demand for memory chips, transforming these components from mere hardware necessities into vital pillars of our global digital infrastructure.
Related articles:
