The ledger does not lie, only the interpreters do. The interpretation from Seoul is clear: the memory cycle is dead. Over the past two weeks, a coordinated wave of target price cuts from major Korean brokerages—including Kiwoom, Mirae Asset, Shinhan Investment, and Samsung Securities—has hit the two pillars of the global memory industry, Samsung Electronics and SK Hynix. The cuts range from 4.5% to a staggering 30%, depending on the analyst. The market is now pricing in a structural peak-out for the memory super-cycle, a narrative that demands a forensic dissection of the underlying assumptions, not a blind acceptance of the sell-side consensus. Let us audit the ledger.
Context: The Architecture of a Cyclical Giant
Samsung Electronics and SK Hynix are not just companies; they are the gatekeepers of the global data storage infrastructure. As IDMs, they control the entire value chain from design to fabrication to packaging. Together, they command roughly 70% of the global DRAM market and 55% of the NAND Flash market. Their primary competitors are Micron Technology in the US, and long-term potential threats from Chinese manufacturers like YMTC and CXMT, who are currently constrained by US export controls on advanced equipment.
The current cycle has been a classic boom-bust-boom narrative. The post-pandemic correction in 2022-2023 saw DRAM and NAND prices collapse, forcing both companies into cost-cutting and production cuts. The recovery in 2024 was driven by two factors: the artificial scarcity created by production cuts, and the explosive demand for High Bandwidth Memory (HBM) from AI data centers. This led to a sharp price recovery, particularly for DDR5 and HBM. The market’s euphoria peaked in mid-2024, with SK Hynix’s stock price hitting record highs. The current sell-off and target price cuts are the market’s attempt to value the next phase of the cycle, which looks markedly different from the previous one.
Core: The Systematic Teardown of the Cycle Thesis
The core of the brokerages’ argument rests on a simple, brutal cycle: supply and demand. However, the specific mechanics of this cycle are more complex than a simple inventory correction. The primary risk is not a sudden collapse in demand, but a structural shift in the composition of that demand, leading to a bifurcation of the market.
1. The HBM vs. Legacy Storage Divergence. The most significant structural change is the split between AI-driven HBM and legacy storage. The brokerages are not betting against HBM. They are betting that the high prices of legacy DRAM (DDR4, DDR5) and NAND are unsustainable. The AI boom has created a massive demand for HBM, which is a high-value, high-margin product. However, HBM production consumes valuable DRAM wafer capacity. By shifting capacity to HBM, both Samsung and SK Hynix have effectively reduced the supply of legacy DRAM, artificially inflating its price. This is a temporary, self-correcting mechanism. The bearish thesis is that the pricing power for legacy DRAM will revert to the mean as the supply constraints from HBM rebalance. The report from Kiwoom, which cut its SK Hynix target by only 4.5%, implicitly acknowledges the strength of HBM, while the more aggressive cuts from Mirae Asset (implied ~30% on Samsung) reflect a deeper concern about the earnings power of the legacy business, which still constitutes the majority of revenue for Samsung.
2. The Capital Expenditure Trap. Memory is a high fixed-cost industry. Capital expenditure as a percentage of revenue can reach 30-40% during expansion phases. During the 2021-2022 boom, both companies committed massive capex. The depreciation from these investments is now a fixed cost. The critical hidden variable is the depreciation schedule. If memory prices fall, the combination of declining revenue and non-decreasing depreciation creates a “double-hit” on profit margins. The brokerages’ model likely assumes that the depreciation from the 2021-2022 capex cycle will be a significant drag on earnings in the second half of 2025, even if HBM volumes remain strong. This is a classic financial engineering problem: revenue volatility is amplified by operational leverage.
3. The Inventory Cycle Re-accumulation. The market is now in the late stages of a “passive inventory depletion” phase. The 2023 production cuts forced downstream customers to draw down their own inventories. The 2024 price recovery triggered a wave of “active inventory replenishment.” The brokerages are forecasting that this replenishment cycle is ending. The next phase is likely to be “active inventory re-accumulation,” where customers, expecting prices to fall, will reduce their orders. This is the classic signal of a cycle peak. The 30% cut in target prices implies a significant de-rating of the forward P/E and P/B multiples, as the market prices in a transition from a “growth-to-price” cycle to a “volume-to-price” cycle.
4. The Financial Engineering of the 30% Cut. The 30% cut is not a 30% cut in fair value. It is a 30% cut in the target price, which is a forward-looking estimate. The report’s data suggests that the underlying earnings forecast for some of these stocks may have been cut by 40-50%. Memory stocks have high beta and high operating leverage. A 10% decline in average selling prices (ASP) can lead to a 30-40% decline in operating profit. The brokerages are not just forecasting a decline in prices; they are forecasting a decline in the rate of change of prices. This is a subtle but critical distinction. The market is not pricing in a crash; it is pricing in a plateau and a subsequent decline. The 30% cut is a mathematical consequence of a model that assumes a realistic, non-linear decline in ASPs, not a panic-driven sell-off.
Contrarian: What the Bulls Got Right
The bearish narrative is compelling, but it is not airtight. The bulls have a strong case, and the brokerages may be underestimating the structural shift in the demand profile.
The HBM Demand Structure is Real. The AI capex cycle is not a speculative bubble. The hyperscalers (Microsoft, Google, Amazon, Meta) are investing billions of dollars in AI infrastructure. The demand for HBM is contractual and long-term. SK Hynix is essentially locked in with NVIDIA for the HBM3E generation. This is not a spot market; it is a negotiated market. The bull case is that the price decline in legacy memory will be offset by the volume and margin growth in HBM and advanced packaging. The brokerages may be incorrectly applying a single-cycle model to a dual-market structure.
The High Barrier to Entry is a Moat. The storage industry is not easy to disrupt. The capital requirements, the process technology knowledge, and the customer certification cycles are immense. The competition from Chinese manufacturers is a long-term threat, but the US export controls on advanced equipment (EUV, high-NA EUV) have created a significant technological moat for the Korean and US players. The brokerages’ bearish thesis does not account for the long-term pricing power of the oligopoly, which can coordinate production cuts to defend prices.
The Valuation is Already Compressed. The current P/B ratios for Samsung and SK Hynix are not at bubble levels. Samsung’s P/B is around 1.2x, and SK Hynix’s is around 2.0x. These are not high multiples by historical standards. The bull case is that the market has already priced in a significant amount of bad news, and the 30% target price cut is a “buy the dip” opportunity for long-term holders. The data shows that the market is not pricing in a full-blown recession; it is pricing in a normalization. The contrarian view is that the brokerages are being too pessimistic about the speed of the normalization.
Takeaway: The Accountability Call
The market is now facing a critical test of its faith in the AI narrative. The 30% target price cuts are not a signal of a systemic collapse. They are a rational, technical, and mathematically driven reassessment of the memory cycle. The question is not whether the cycle will peak, but when, and how fast. The brokerages are betting on a fast de-rating. The bulls are betting on a slow, structural shift. The data supports the bears in the short term, but the long-term ledger is still being written. History repeats, but the gas fees change. The real risk is not the cycle itself, but the assumption that the cycle will behave exactly as it did in the past. Trust is a bug, not a feature. Verify the price action, ignore the narrative. The crunch is coming. The only question is the landing.