Press Enter to search · ESC to close

US Stocks

Memory Stocks at 2-3x 2027-28E P/E: A Cyclical Bottom, Not a Downturn, Says BofA

Bank of America argues that SK Hynix and Samsung’s 2-3x 2027-28E P/E reflects a cyclical bottom, not a structural downturn, as AI memory demand remains strong and Q4 pricing holds up.

Memory stocks: deep correction, but fundamentals intact

SK Hynix and Samsung Electronics have fallen more than 50% from their June peaks, leaving them trading at just 2-3x consensus 2027-2028 earnings per share. In an October 16, 2026 report, Bank of America placed these valuations in cyclical-bottom territory while maintaining super-cycle-level earnings forecasts. The bank argues that the recent correction is disconnected from memory fundamentals and reflects excessive fears of an AI memory boom-bust.

Two-layer valuation framework

BofA’s analysis splits into valuation and earnings layers. On valuation, a 2-3x P/E is well below the 4-7x seen in previous upcycles, and price-to-book is near 1.0x 2027-28E book value. Historically, such levels appear at cycle troughs, typically accompanied by high chip inventories, DRAM average selling prices down 50% or more, large operating losses, and high borrowing costs. Today, DRAM ASPs are up 20-30% year-over-year—the opposite of historical bottom conditions.

On earnings, BofA acknowledges softer 2028 numbers: single-digit EPS decline year-over-year, with sequential growth slowing in 2027. Yet absolute sales, operating profit, and EPS remain near 2027 peak and record levels. Even using the softer 2028 earnings, valuations look cheap. Samsung trades at 3.0x 2027E and 2.8x 2028E P/E; SK Hynix at 2.7x and 2.3x, respectively.

AI memory demand still strong

Data center AI memory replacement demand spans LPDDR5, SOCAMM, HBM, and eSSD. More advanced GPUs and ASICs continue to raise memory content per system. SK Hynix’s current correction—down 57% from June to October—exceeds its 20-year average drawdown of 51%, but it happened in just four months versus the historical 15 months. The speed and depth suggest a bottom formation rather than a structural downturn.

Investor debate: direction agreed, cycle length disputed

BofA met investors in Singapore this week. Hedge funds are bearish, while most long-only investors are bullish. Bears cite high borrowing costs for neoclouds, rising Chinese 4F² DRAM supply, declining memory content per GPU, slowing AI capex or memory price growth, and earnings missing bullish consensus. BofA rebuts each point: valuations are too low even with softer 2028 earnings; China’s 4F² impact is limited; AI startups and neoclouds have large backlogs; memory makers can maintain high margins and ROIC even with softer ASPs after 2026; shareholder returns could exceed 10% of current market cap via dividends and buybacks in 2027-28.

Q4 guidance and pricing remain strong

Nanya Technology has issued bullish Q4 guidance. BofA expects SK Hynix’s October 28 Q4 guidance to show rising ASPs and margins. HBM4 12-layer ramp is strong, with additional 8-layer output for new orders; a special cash dividend is possible despite the recent 40 trillion won buyback. October industry data is not weak: PC DRAM contract prices for DDR5 and DDR4 rose over 15% month-over-month; spot prices held stable despite rising US Treasury yields; South Korea’s semiconductor exports for the first 10 days of October were $12.2 billion, down 26% month-over-month but near record highs; China’s August memory imports reached $37.3 billion, 57% of total imports, up 277% year-over-year. September China IC imports hit $71.7 billion, up 75% year-over-year, exceeding $70 billion for the first time.

Forward look

BofA’s conclusion: the deep correction looks like a cyclical bottom, but has not turned into a downturn. Memory stock valuations are already pricing a cycle trough. Whether Q4 guidance can pull fundamental confirmation back to the upside is the key variable to watch. For crypto and AI infrastructure investors, memory is a critical input for GPU clusters and decentralized compute networks—sustained memory demand supports the buildout of AI infrastructure that increasingly intersects with blockchain-based compute markets.

View original

Share
Risk notice This site provides news and information on the crypto, blockchain and Web3 industry for reference only and does not constitute investment advice or any promise of returns. Virtual currency-related activities are illegal financial activities in mainland China; digital asset prices are highly volatile; use at your own risk. This site does not provide trading, token issuance or related referral services.

Related Reading

Latest News

TREE NEWS share card
Long-press image above → Save to Photos / Share
Pitch us Feedback