SK Hynix expects the supply squeeze in AI-related memory chips to last until the end of 2030, Chief Executive Kwak Noh-jung said, offering an unusually long outlook for the sector. The view suggested that the current memory upcycle could be supported by more predictable AI-driven demand than in previous semiconductor cycles.
Kwak made the comments on August 27 after a groundbreaking event for SK Hynix’s advanced AI memory packaging facility at Purdue University in West Lafayette, Indiana.
Addressing questions about a possible oversupply phase and a future industry downturn, Kwak said the exact duration of the shortage cannot be known. However, he said there are currently no obvious signals that the cycle is turning, and he expects the present environment to remain in place through 2030.
Kwak said any future downturn would likely differ from past memory-market contractions. Historically, memory chips were largely standardized products, with manufacturers increasing production for the same broad market. Because demand was difficult to measure precisely, companies often struggled to align output with actual customer needs.
That pattern is changing as AI reshapes memory requirements, he said. High-bandwidth memory, or HBM, is increasingly designed to work with AI processors such as GPUs, tying production more closely to individual customer specifications.
According to Kwak, the industry is moving away from fully standardized products toward partially or fully customized memory solutions. He said next-generation products now require closer development work with customers to improve system performance, which allows suppliers to estimate demand with greater accuracy.
As a result, Kwak said a future downturn may not resemble the steep declines seen in earlier decades. Instead, he expects any weakening to be more gradual or to resemble a flat market trend rather than a sharp contraction.
The comments represent one of SK Hynix’s clearest long-range assessments of the current memory expansion. Kwak said AI business models are reducing the gap between supply and specific demand, limiting the potential for excessive inventory build-ups.
Looking beyond 2030, Kwak said he expects supply and demand to move toward equilibrium while the broader AI industry continues developing.
