Memory chip shortage may get worse in 2027 as HBM demand squeezes DRAM supply
The pressure on the global memory chip market may be far from over. While consumers and hardware makers are already dealing with rising component costs and tight supply, SK hynix CEO Kwak Noh-jung has warned that the industry could face an even more difficult period ahead.
According to Kwak, 2027 may become the toughest year in the memory industry’s history from a supply standpoint. He also indicated that customer demand could continue to exceed available supply capacity even beyond 2030, suggesting that the current imbalance may not be a short-term problem.
A key reason behind this pressure is the explosive growth of high-bandwidth memory, better known as HBM. Demand for HBM has surged as artificial intelligence servers, data centers, and advanced GPUs require faster and more efficient memory. This shift is reshaping how manufacturers allocate production capacity.
Industry estimates suggest that memory-related wafer supply could grow by around 12% annually in 2027 and 2028. However, roughly half of that added capacity is expected to be absorbed by HBM production. That leaves much less room for traditional DRAM supply growth, even as demand for conventional memory remains strong across PCs, smartphones, servers, and consumer electronics.
As a result, non-HBM DRAM bit growth may be limited to about 15% per year during 2027 and 2028. That would fall short of expected demand growth of around 22%, creating a gap that could keep prices elevated and supply tight.
Samsung is expected to ship around 12 billion Gb of HBM in 2026, increasing to approximately 20 billion Gb in 2027. SK hynix is projected to supply about 18 billion Gb of HBM in 2026, rising to 24 billion Gb in 2027. These figures show just how aggressively major memory makers are shifting capacity toward AI-focused memory products.
New fabrication plants should help increase output, but they will not arrive quickly enough to fully ease the pressure. Samsung’s P5 Fab 1 is expected to begin operations by July 2027, while P5 Fab 2 and another facility in Yongin are planned for production by 2029. SK hynix’s Yongin Y1 fab is expected to come online in February 2027, followed by Y2 in the second half of 2028.
Even with these expansion plans, the memory market may remain constrained because building advanced semiconductor capacity takes years. At the same time, AI demand is growing at a pace that continues to challenge supply forecasts.
Another potential factor is CXMT’s planned move toward DDR6 memory. If successful, this could help ease some supply pressure in the longer term. However, production timelines remain unclear, and it is still uncertain how much supply the company can provide outside China.
For consumers, the result could be continued price pressure across electronics that rely on DRAM and NAND components. PCs, laptops, graphics cards, smartphones, gaming devices, and servers may all feel the impact if memory prices remain high.
For the broader technology industry, the message is clear: the AI boom is not only driving demand for GPUs and accelerators, but also fundamentally reshaping the memory supply chain. HBM is becoming one of the most important components in the race to build more powerful AI infrastructure, and that demand may keep the entire memory market under strain for years.
If current forecasts prove accurate, 2027 could mark one of the most challenging periods yet for global memory supply, with the effects likely extending well into the next decade.
