The relentless demand for AI infrastructure is pushing the leading players in the semiconductor sector to revise upward their industrial strategies.
SK Hynix, currently the world’s largest supplier of HBM memories used in high-end graphics cards, plans to triple its production capacity.
However, the fruits of this massive expansion will not be visible immediately. As confirmed by the president of SK Group, Chey Tae-won, in a recent interview with the publication Nikkei Asia, the final milestone will be reached only in 2034.
SK Hynix wants to triple production, but it will take a very long time
This statement follows shortly after the company’s announcement of doubling production within the next five years. The management’s revised calculations indicate that, once the plants currently under construction are completed, the volume of silicon wafers produced will undergo an increase well beyond initial estimates, thereby tripling current volumes.
The company is indeed accelerating the activation of four new factories, whose first phase is expected to become operational starting in 2027.
Although the original plans envisioned a development of the sites spread over two decades, the timelines have been drastically compressed to try to meet the market’s chip hunger.
However, President Chey has clarified that there is no feasible way to proceed faster, given the stringent technical constraints associated with building cutting-edge facilities. Assembling and starting up a next-generation memory factory requires a minimum of about 3 years.
The current dynamics have generated a substantial increase in SK Hynix’s corporate value, which, together with assieme a Samsung and Micron, firmly controls the production of NAND and DRAM memories.
These components are essential to support the growing workloads related to artificial intelligence, but this immense demand has triggered a steep cascade of price increases, hitting end consumers hard.
The costs of DRAM modules and SSD units have soared in recent months, with some products having more than tripled in value compared to the same period a year earlier, delivering record revenues for manufacturers.
Despite the planned increase in production capacity, buyers should not expect cost relief in the short term. Market analyses indicate that the price of memory will reach its highest peak no earlier than the end of the current year, then stabilize around 2027, avoiding the rapid downturns seen in past market cycles.
International expansion and market limits
In addition to consolidating its presence on domestic soil in South Korea, the tech giant is exploring new options for manufacturing abroad. Japan emerges as an excellent candidate to host future production sites, backed by an extremely solid supply chain in the semiconductor sector.
The usual fluctuating economic cycle of electronic components, historically characterized by alternating phases of severe shortage and saturation, has undergone a profound alteration.
The explosion of AI-related demand that began in 2022 coincided with a period of deep financial contraction for the DRAM industry.
This unfavorable timing has made it particularly costly to raise the capital needed to start the new sites, highlighting the rigidity of an industry forced to contend with colossal investments and decade-long timelines.
