Farewell to the cycle of fate! SK Hynix (SKHY.US) CEO warns that the shortage of storage chips will continue for the next decade. The group chairman introduces a new concept of "memory as a service".

Against the backdrop of a continuous surge in demand for artificial intelligence (AI) computing power, global storage chip giant SK Hynix has just completed a historic listing in the United States. At the same time, senior executives of the company have also made a series of heavyweight statements: they not only predict that 2027 will see the most severe supply shortage in industry history, but also reveal that the company is planning to revolutionize the traditional hardware sales model with a "memory as a service" business model. These statements not only show the extreme optimism of this core AI supply chain enterprise about the future, but also attempt to completely rewrite the "boom-bust" cycle that has plagued the storage chip industry for decades. Supply crisis in 2027: Demand will "overwhelm" production capacity for a long time SK Hynix CEO Kwak Noh-jung issued the most radical shortage warning in the industry so far in an interview. He clearly stated that the global storage chip industry is heading towards the "most severe supply shortage in history" by 2027, and this imbalance of demand far exceeding supply will persist throughout the next decade. "We predict that from the supply perspective, next year will be the worst year in the industry's history," Kwak Noh-jung pointed out on the day SK Hynix's American Depositary Receipts (ADR) started trading on Nasdaq. He explained that despite the company's aggressive capacity expansion, customer demand continues to rise, and manufacturing capacity has physical limits. "We still predict that even after 2030, customer demand will still exceed our supply capacity. But we are doing our best to address this issue." SK Group Chairman Chey Tae-won, in an interview on the same day, described the supply bottleneck in a more exaggerated dimension. He believes that the supply and demand of storage chips may not normalize until humans achieve "general artificial intelligence" (AGI). "Until that day comes, we need a lot of memory." Chey Tae-won stated that the arrival of the ChatGPT era has triggered a continuous and widely spread memory shortage that is driving up the prices of various electronic products from iPads to Xbox gaming consoles through the supply chain. Breaking the curse of cycles: Signing long-term contracts and evolving the business model towards "memory as a service" For a long time, the memory chip industry has been plagued by severe "boom-bust" cycles, with manufacturers often overproducing during prosperous times and then facing price collapses. However, Kwak Noh-jung asserted that "the situation has clearly changed." He pointed out that customers no longer make short-term spot purchases as they did in the past but actively seek to sign long-term supply agreements. "They believe that the shortage situation will continue for a longer time," stated Kwak Noh-jung. Chey Tae-won was more straightforward in declaring, "This is no longer a cyclical business." He explained that even in the event of an industry downturn in the future, these long-term agreements can help the company maintain shipments and memory chip prices, "This actually puts us in a completely different situation." It is worth noting that Chey Tae-won revealed that SK Hynix is contemplating a new business model - "Memory as a Service" (Memory as a Service). This means that in the future, customers may no longer need to directly purchase physical semiconductors but can "lease" memory usage rights from SK Hynix. Chey Tae-won said, "We can become a storage service provider. In the future, this is another area we can focus on." Although he did not elaborate on how this model will work, he emphasized that the goal is to address the bottleneck of memory capacity and bluntly stated, "We must overcome this challenge." This concept is similar to the SaaS model in the software industry - where customers can pay to use software or computing power without owning it outright. US factory suspense resurges, investment commitment far exceeds $35 billion Regarding global production capacity layout, Kwak Noh-jung stated that the United States is still one of the candidate sites for future wafer fab investments, but no final decision has been made yet. He pointed out that the core criteria for site selection are whether competitive manufacturing costs can be achieved, as well as the availability of sufficient land, power, water resources, and skilled workers. If the conditions are met, the United States, Japan, and Southeast Asia are all under consideration. At the same time, the SK Group is making a huge bet on the United States. Chey Tae-won revealed that the group's planned investment in the United States far exceeds the currently announced $35 billion, "My plan is a much larger figure, far exceeding $35 billion." Currently, SK Hynix has committed to investing about $4 billion to build an advanced chip packaging factory in Indiana and another $10 billion to develop AI solution companies in the US to seek new AI growth engines. Regarding the Nasdaq listing, Chey Tae-won remains open-minded, hinting that there may be further stock issuance in the United States in the future, but the premise is to bring substantial returns to new investors first. "What we have to do first is to keep the stock price stable, and in the long run, we hope to unleash more upward potential." Divergence in AI investment feast, industry and capital cautiously playing games Despite the AI investment frenzy driving up the entire semiconductor sector, concerns about a bubble have never dissipated in the market. Recent reports indicate that Apple Inc. (AAPL.US) is seeking to diversify parts of its semiconductor supply chain to Chinese suppliers, while Meta (META.US) is attempting to commercialize excess AI computing power. These reports have sparked investor concerns about an impending turning point in the AI investment cycle. Wall Street heavyweights such as Michael Burry and Ray Dalio, famous for their bearish views, have also warned that the AI bubble will eventually burst. However, the industry and some investment banks remain bullish. NVIDIA Corporation (NVDA.US) CEO Jensen Huang stated last month that due to strong demand, the AI memory shortage will persist for several years, confirming that SK Hynix remains its largest memory supplier. UBS Group AG also forecasts that the global DRAM industry will remain undersupplied at least until the second quarter of 2028. Bank of America continues to be bullish on the AI investment cycle, forecasting that global mega-scale cloud companies' capital expenditures will reach approximately $851 billion this year and soar to $1.15 trillion next year. The bank believes that the approximately $244 billion raised by cloud companies this year is more for balance sheet optimization rather than a signal of financial stress. Micron Technology, Inc. has previously raised its investment plan in the US from $200 billion to $250 billion by 2035, further confirming the industry's fervent expectations for memory demand in the AI era. From the bottom to the peak: A roller coaster ride of stock prices and performance Thanks to its heavy investment in the forward-looking High Bandwidth Memory (HBM), SK Hynix has become one of the biggest winners in the AI gold rush. The company's operating profit reached a record 470 trillion won (approximately $31 billion) in 2025, doubling from 2024 and completely reversing the loss situation in 2023. This growth trajectory further accelerated in 2026, with the upcoming quarter from April to June expected to reach a staggering 655 trillion won in operating profit, according to LSEG data. The capital market has also reacted dramatically. Despite a recent 18% pullback in stock prices due to doubts about the sustainability of the AI rally in the past two weeks, SK Hynix's stock price has soared over seven times in the past 12 months. On the first day of its Nasdaq listing, its ADR price surged by 12.76% to close at $168.01, completing the largest initial public offering by a foreign company in the US market in history, raising $26.5 billion. Kwak Noh-jung likened the current AI infrastructure construction to the expansion of the Internet, believing that this process will continue for decades. "It took us almost 30 years to complete the construction of the Internet infrastructure. As for AI, I believe its industry will be much larger than the Internet."