The memory chip industry is approaching a rare historical inflection point. According to UBS's latest research report, AI-driven demand for High Bandwidth Memory (HBM) continues to encroach on DDR production capacity. Combined with the simultaneous surge from the traditional server replacement cycle and solid-state drive (SSD) demand, the global DRAM market's supply-demand gap is projected to extend into the fourth quarter of 2027. This represents a memory supercycle not witnessed in nearly thirty years.
Furthermore, the DRAM market has consolidated into a tight oligopoly dominated by three major players. The supply side lacks sufficient incentive for "competitive capacity expansion" to mitigate demand shocks, leading to less severe cyclical corrections. Consequently, the upward price pressure from this demand supercycle is expected to be more persistent.
SK Hynix holds a leadership position in the HBM sector. UBS has raised its 12-month target price for the company from 1.55 million won to 1.70 million won. Concurrently, UBS significantly upgraded its earnings forecasts: EPS estimates for 2026 and 2027 were increased by 22% and 29% respectively. The 2026 operating profit forecast is a substantial 286 trillion won, approximately 57% higher than the market consensus.
Regarding near-term catalysts, SK Hynix's potential ADR listing may be accompanied by a share buyback program in the Korean market. This would create a "dual positive effect"—enhancing accessibility for overseas investors while directly supporting the share price in Korea.
**Underlying Logic of the Once-in-30-Years DRAM Shortage Cycle**
The current memory chip cycle is historically rare, driven by a confluence of multiple demand-side shocks. Firstly, explosive growth in HBM demand is persistently crowding out traditional DDR capacity. By the end of 2026, global front-end HBM-dedicated DRAM capacity is expected to reach 500,000 wafers per month (12-inch equivalent), accounting for 25% of total industry capacity, rising to approximately 31% by 2027.
Secondly, the traditional server replacement cycle is underway, with continued demand release for conventional DRAM like DDR5. Simultaneously, the expansion of AI infrastructure is driving incremental demand for servers and storage SSDs.
Thirdly, nearly all wafer capacity expansion outside China is concentrated on HBM, leaving new supply for conventional DRAM extremely limited. The three major suppliers show significantly constrained willingness to expand capacity in non-HBM areas.
The highly consolidated oligopolistic structure of the DRAM market means reduced incentive for competitive capacity expansion to meet demand shocks, lessening the severity of cyclical corrections and leading to more persistent upward price pressure.
**SK Hynix's HBM Leadership: Short-Term Concerns Don't Alter Long-Term Outlook**
Recent market concerns regarding SK Hynix focus on two areas: minor redesign issues for HBM4 products and HBM shipment volumes. Neither is considered a fundamental threat to its leading position.
Regarding the HBM4 redesign, SK Hynix is reportedly finalizing minor adjustments for the logic die and DRAM die and progressing towards final certification for NVIDIA's Rubin platform. Consequently, market share assumptions for HBM4/Rubin are adjusted to 60% for SK Hynix, 30% for Samsung, and 10% for Micron.
For shipments, SK Hynix's HBM bit output is forecast at 18.4 billion Gb in 2026 (a 46% year-on-year increase) and 24.7 billion Gb in 2027 (a 34% increase). In terms of industry bit share, SK Hynix is projected to maintain its top position in the HBM market for both years, with shares of 51% and 44% respectively.
The technological expertise and customer trust built through SK Hynix's execution capability in HBM over recent years form a core barrier sustaining its leadership. The HBM4 design adjustments are viewed as optimizations rather than a directional shift.
**Earnings Significantly Outpacing Expectations: Substantial Valuation Upside**
Based on this analysis, UBS's earnings forecasts for SK Hynix substantially exceed market consensus. For revenue, 2026 is forecast at 355.1 trillion won (approximately 41% above the consensus of 251.6 trillion won), and 2027 at 531.6 trillion won (approximately 65% above the consensus of 323 trillion won).
For operating profit, the 2026 forecast is 286 trillion won (about 57% above the consensus of 182 trillion won), and 443.5 trillion won for 2027 (about 88% above the consensus of 235.6 trillion won). Specifically for Q1 2026, the operating profit forecast is 41.76 trillion won, significantly higher than the consensus estimate of 32.2 trillion won.
Regarding cash flow, free cash flow is projected to be 143.8 trillion won in 2026, surging further to 269.8 trillion won in 2027, corresponding to free cash flow yields of approximately 21.6% and 40.5% respectively.
**Valuation and Catalysts: ADR Listing and Buybacks Offer Dual Support**
Using a 12-month forward price-to-book (P/B) multiple valuation method, and based on a long-term ROE forecast of 32.1% and a cost of equity of 11.2%, a target P/B ratio of 2.86x is derived, corresponding to the target price of 1.70 million won.
The current Next-Twelve-Months (NTM) P/B ratio of 1.54x implies a long-term ROE of only 17.3%. This is not only below the historical 10-year average of 21% but also significantly lower than UBS's forecast of 32.1%.
This valuation gap suggests the market has not yet fully priced in SK Hynix's higher DRAM profitability and strong free cash flow generation capabilities.
As near-term catalysts, the potential ADR listing, possibly coupled with a domestic share buyback, presents a dual positive. Additionally, around April 24, 2026, positive catalysts are anticipated, including smartphone sales data, pricing negotiation outcomes, and GPU supply chain dynamics.
Comments