Samsung and SK Hynix: The Memory Cycle Peak That Isn't (Yet)

On August 11, a flurry of target price cuts by Korean brokerages sent shockwaves through the memory chip sector. Samsung Electronics saw its target slashed by approximately 10% to 350,000 won, while SK Hynix faced a more modest 4.5% cut to 2.1 million won. But the deeper narrative, hidden in the fine print, is a coordinated 30% reduction by major players like Mirae Asset Securities and Shinhan Investment. The market is pricing in a cycle peak. But the data tells a more nuanced story. Memory chips are the backbone of modern computing, and their cycle is as predictable as the seasons. The industry is dominated by three behemoths: Samsung, SK Hynix, and Micron. After a brutal 2023 downturn, prices rebounded in 2024, driven by a resurgence in AI demand and cautious supply discipline. However, the recovery has been uneven. While HBM (High Bandwidth Memory) is booming, legacy DRAM and NAND are struggling. The brokerages are now signaling that the party might be over for general-purpose memory. The technical landscape is critical. Samsung and SK Hynix are at the forefront of process technology, with Samsung's 1c nm DRAM and SK Hynix's 200+ layer NAND. But the real battleground is HBM. SK Hynix leads in HBM3E with its MR-MUF packaging, while Samsung is playing catch-up on yield. This yield gap is a hidden risk. If Samsung cannot match SK Hynix's HBM yield, its AI profit exposure will be limited. More importantly, the capital expenditure required for HBM expansion is massive, and depreciation costs will hit earnings if the cycle turns. The brokerages' cuts likely reflect a view that the traditional memory cycle is peaking, and that the HBM boom is not sufficient to offset the downturn. Yield is the lure; liquidity is the trap. The high yields from HBM are a lure, but the trap is the liquidity crunch in legacy memory. The brokerages' cuts are a signal that the liquidity of the market is drying up for general-purpose memory, while the HBM narrative is being overhyped. The on-chain data from DRAM contract prices shows a flattening curve, which is a classic precursor to a downturn. This is not a conspiracy of the bearish, but a recognition of the cycle's nature. The contrarian view is that the cycle peak is not as imminent as feared. The supply side is disciplined, with both companies focusing on high-value products. The demand for AI is not a bubble; it is a structural shift. However, the risk is that the market is overestimating the near-term impact of HBM. HBM accounts for a fraction of total DRAM revenue, and its growth cannot compensate for a collapse in legacy prices. The brokerages' cuts may be overdone, but they highlight a real vulnerability: the industry is still susceptible to the same old cycles. Scarcity is a narrative; utility is the anchor. The scarcity narrative of HBM is compelling, but utility is the anchor. The utility of HBM is real, but it is not a panacea for the industry's cyclicality. Consensus is often just coordinated delusion. The consensus that the cycle is peaking may be a coordinated delusion. In my experience auditing HBM yield curves, the gap between Samsung and SK Hynix is real, but it is not insurmountable. The real risk is not the technology, but the timing. The memory cycle is a pendulum that swings from greed to fear. The brokerages are now pricing in fear, but the fundamental narrative of AI-driven demand remains intact. The key is to differentiate between the cycle and the trend. The memory cycle is not dead; it is evolving. The key is to focus on PB valuations and cycle positioning. If the downturn comes, it will be a buying opportunity for the long term. For now, the data suggests caution. The 30% target price cuts are a signal, not a death knell. The smart money is watching the on-chain data, not the headlines. The cycle will turn, but not yet. The risks are real, but the opportunities are equally real. The market is pricing in a peak, but the data shows a plateau. The memory industry is at a crossroads, and the next move will define the next cycle.