Memory stocks hit cycle-bottom valuations, but fundamentals haven’t rolled over
TREE NEWS reports: South Korean memory giants SK Hynix and Samsung Electronics have suffered a brutal selloff, with SK Hynix down 57% from its June peak and Samsung off more than 50%. Yet a Bank of America research note dated October 16, 2026 argues that the stocks now trade at just 2-3x 2027-2028 consensus earnings per share and near 1.0x 2027-28 estimated book value—levels that scream cycle bottom rather than fundamental deterioration.
The bank maintains its super-cycle forecasts for average selling prices, revenue, and margins, projecting that 2028 sales, operating profit, and EPS will remain close to 2027 peaks. That view stands in sharp contrast to the market’s fear that the memory upcycle is over.
Why the disconnect matters
Memory is the most cyclical corner of semiconductors, and investors have historically fled at the first sign of peak pricing. But BofA’s argument is that this cycle is structurally different: AI-driven demand for high-bandwidth memory (HBM) has created a supply-demand imbalance that extends beyond traditional PC and smartphone cycles. The bank expects SK Hynix’s fourth-quarter guidance on October 28 to show rising ASPs and margins, with strong ramp-up of 12-layer HBM4 production.
If that guidance confirms, it could mark a turning point for memory stocks—and by extension, the broader semiconductor complex. The valuation math is striking: at 2-3x forward earnings, investors are pricing in a severe downcycle that BofA says isn’t materializing.
What to watch
- October 28: SK Hynix Q4 guidance—ASPs, margins, and HBM4 ramp commentary.
- Pricing data: Contract prices for DRAM and NAND in Q4 and early 2027.
- HBM4 qualification: Customer certifications for 12-layer HBM4, a key margin driver.
- Inventory levels: Whether channel inventories remain lean despite production increases.
For crypto markets, the read-through is indirect but relevant. Memory chips power the AI data centers that underpin demand for GPU-based compute networks and decentralized AI infrastructure. A healthy memory cycle supports capex spending across the tech stack, including blockchain-based compute marketplaces. Conversely, if BofA is wrong and memory rolls over, risk appetite for AI-adjacent crypto assets could suffer.
Forward look
The next two weeks are critical. If SK Hynix’s guidance confirms upward ASP and margin trends, the 2-3x P/E argument will gain traction and could trigger a sharp re-rating. If guidance disappoints, the market’s pessimism will be validated. Either way, the memory cycle remains a key bellwether for the AI trade—and for the crypto tokens levered to AI compute demand.
