SK hynix Plans Major DRAM Expansion as AI Demand Pushes Memory Supply to the Limit
SK hynix is preparing for one of the most ambitious memory production expansions in its history, with plans to nearly double its DRAM wafer capacity by the end of the decade. The move comes as demand for artificial intelligence hardware continues to surge, putting pressure not only on GPU production but also on the high-performance memory needed to power modern AI accelerators.
According to supply chain information from South Korea, SK hynix had already been working on a large-scale capacity expansion before NVIDIA CEO Jensen Huang publicly urged the company to “please make more” during Computex in Taiwan. Huang reportedly wrote the message on a SK hynix wafer, highlighting just how critical memory supply has become for the AI industry.
The company is said to be aiming for a monthly DRAM wafer input capacity of around one million wafers by 2030, up from its current level of roughly 550,000 wafers per month. This would represent a major increase in global DRAM output and could help ease bottlenecks in the production of high bandwidth memory, better known as HBM.
HBM has become one of the most important components inside AI GPUs. While much of the spotlight has been on advanced chips and packaging technologies, memory has quietly become a key constraint in the AI hardware supply chain. AI accelerators require massive memory bandwidth to handle large language models, training workloads, inference tasks, and data center applications. As a result, demand for HBM has climbed sharply alongside the rapid rollout of AI infrastructure.
At the heart of SK hynix’s expansion strategy is its Yongin semiconductor cluster in South Korea. The facility is expected to play a central role in increasing production over the next several years. Equipment installation for the first cleanroom is reportedly scheduled to begin in February 2027, with capacity additions expected to continue in stages.
The first phase of the Yongin site is planned around six cleanrooms. SK hynix is expected to add around 60,000 wafers of monthly capacity during the initial stage, followed by similar capacity increases every six months. If the schedule proceeds as planned, the first phase alone could reach approximately 360,000 wafers per month by 2030.
The company is also expected to expand production at its Cheongju Technopolis complex, where its MI5X facility could receive an additional 80,000 wafers of monthly capacity. Together, these investments would significantly strengthen SK hynix’s ability to supply advanced DRAM products for AI, data centers, servers, and next-generation computing devices.
SK hynix’s Wuxi plant in China is also expected to remain important in the company’s production strategy. The site currently accounts for a large portion of the company’s DRAM output and could support the broader effort to scale supply as global memory demand grows.
The planned expansion reflects the changing balance of power in the semiconductor market. In previous cycles, memory manufacturers often faced oversupply risks when demand cooled after aggressive capital spending. This time, however, AI is creating a new layer of long-term demand, especially for premium memory products such as HBM.
Still, some industry observers remain cautious. Large production increases require massive investment in cleanrooms, equipment, materials, and process technology. There are also concerns about whether demand will continue rising fast enough to absorb the additional output once the new capacity comes online. Suppliers and manufacturing partners are reportedly watching closely, as previous investment booms in the memory industry were sometimes followed by sudden slowdowns in equipment orders.
Even with those risks, SK hynix appears to be positioning itself for a future where memory is just as strategically important as processors. NVIDIA’s AI GPUs, and similar accelerators from other chipmakers, depend heavily on advanced memory to deliver the performance required by modern AI workloads. Without enough HBM and DRAM capacity, the wider AI hardware market could face supply limits.
By targeting one million DRAM wafers per month by 2030, SK hynix is making a clear bet that artificial intelligence, cloud computing, and high-performance data centers will continue driving demand for advanced memory. If successful, the expansion could strengthen the company’s position as one of the leading suppliers in the global AI semiconductor supply chain.
