SK Hynix wants to significantly scale up production of AI memory chips to respond to rapidly growing demand from data centers worldwide. This was announced by Chey Tae-won, chairman of SK Group, the parent company of the South Korean chip manufacturer.
This was reported by Bloomberg. The global wave of investment in AI infrastructure is putting the memory sector under considerable pressure. Data centers need more and more specialized memory chips to train and run AI models, with high-bandwidth memory playing an increasingly important role.
According to Chey Tae-won, capital investments in 2026 will be significantly higher than in previous years. The additional investments are necessary to be able to supply sufficient HBM chips, which are essential for AI accelerators such as those developed by Nvidia. These chips are used to train and run large-scale AI models and are an indispensable part of modern data centers.
During a conference in Washington, the CEO called high-bandwidth memory an exceptionally profitable product. Strong demand for this type of memory has led to SK Hynix’s market value skyrocketing over the past year. The company’s share price has more than quadrupled during that period, partly thanks to record results.
Chey did not specify the exact size of the planned capacity expansion. However, SK Hynix had previously announced that investments would increase significantly in 2026 in order to meet demand. Pressure on the market is high, as major US technology companies are collectively allocating approximately $650 billion for AI-related infrastructure this year. Companies such as Microsoft and Meta are investing heavily to gain a strategic advantage in the competition surrounding artificial intelligence.
Major shortage of memory chips
This urge to invest has now led to a global shortage of memory chips. The market is dominated by three players: SK Hynix, Samsung Electronics, and Micron Technology. SK Hynix has already sold its entire production capacity for memory chips in 2026, while Micron has achieved the same with its HBM products.
Despite the current economic boom, Chey tempered the optimism. He pointed out that rapid technological changes could radically alter the competitive landscape, which could also have negative financial consequences. Analysts now expect SK Hynix’s operating profit to average around $70 billion in 2026, up from the approximately $50 billion predicted at the end of last year.
Some estimates even predict more than $100 billion. However, according to Chey, such a scenario is not guaranteed. He believes that the same dynamics could just as easily lead to the company facing a loss of a similar magnitude.
In addition to chip production itself, SK Group also sees major challenges in the area of infrastructure. The energy supply for AI data centers is becoming an increasingly significant bottleneck. Chey indicated that the group is investigating whether it will be necessary to build power plants in the immediate vicinity of data centers. According to him, the inability to supply sufficient power could have major consequences for further growth in the AI sector.