New Memory Supply May Take Much Longer Than Expected as Samsung and SK Hynix Face Fab Delays
Fresh analysis from Bank of America is challenging the belief that a major wave of new memory chip supply is close to arriving. According to the report, South Korea’s ambitious plan to sharply expand memory production by 2030 may be far more difficult to achieve than government targets suggest.
The outlook is especially important for the global semiconductor market, where demand for DRAM, HBM, and other memory products has surged alongside the rapid growth of artificial intelligence hardware. With AI servers consuming huge amounts of high-bandwidth memory, chipmakers have been racing to expand capacity. However, the latest assessment indicates that meaningful new supply could take many years to materialize.
South Korea’s President Lee Jae-myung has previously expressed a goal of doubling the country’s memory production capacity by 2030. That optimism has been supported by major fabrication projects from Samsung and SK Hynix in the southwestern regions of Gwangju and Jeolla. These sites are expected to become key pillars of South Korea’s future semiconductor manufacturing base.
But industry estimates now suggest the timeline may be too aggressive.
After factoring in the retirement of older production lines, technology transitions, and process miniaturization, South Korea may only be able to increase its active memory wafer capacity by less than 10 percent annually. That pace would leave the country well short of the goal of doubling memory capacity by the end of the decade.
One Taiwanese memory chip industry insider reportedly believes SK Hynix may manage to add only around one-sixth of its originally planned new capacity by 2028. If accurate, that would significantly reduce expectations for near-term supply relief in the memory market.
The biggest challenge appears to be the sheer complexity of building advanced semiconductor fabs from the ground up. Establishing the foundations for the new memory fabrication plants in Gwangju and Jeolla alone could take roughly five years. After that, another three to four years may be needed to install clean rooms, production tools, and other highly specialized chipmaking equipment.
When the broader support ecosystem is included, the full buildout of these sites could take more than a decade. That means the market may not see the full benefits of these investments until well after 2030.
The timing matters because the memory industry is already under pressure from rising prices and tight supply. Samsung, SK Hynix, and Micron dominate the global DRAM market, and their production decisions have a major influence on pricing across consumer electronics, PCs, servers, smartphones, and AI infrastructure.
The situation has also drawn legal scrutiny. Samsung, SK Hynix, and Micron were sued in a California federal court on June 25 over allegations of collusion and price-fixing in the DRAM market. The lawsuit seeks to represent consumers and businesses that purchased products containing commodity DRAM during the recent period of rising prices.
The complaint argues that the three companies used their dominant market positions to shift production toward AI-focused high-bandwidth memory while reducing output of older DRAM types such as DDR3 and DDR4. According to the lawsuit, this alleged coordinated move contributed to higher prices for traditional memory products.
If new memory capacity arrives much slower than expected, it could complicate the industry’s defense. A limited capacity expansion by SK Hynix through 2028 would make it harder to argue that supply constraints will be quickly resolved through new production.
For consumers and businesses, the key takeaway is simple: relief in memory pricing may not come as soon as many expected. AI-driven demand continues to reshape the semiconductor supply chain, and the construction of new fabs is proving to be a long and expensive process.
While Samsung and SK Hynix remain central to South Korea’s semiconductor ambitions, the path to doubling national memory capacity by 2030 now looks increasingly uncertain. The next few years could determine whether the industry can balance booming AI demand with enough DRAM and HBM supply to stabilize prices across the broader technology market.
