HSBC: Samsung Electronics' Stock Has Erased Its Long-Term AI Premium
Claire Weston
HSBC's August 3 report shows Samsung Electronics' implied long-term earnings have fallen to roughly 0.8× its 2024 EPS — meaning the market now prices the company's through-cycle profitability below pre-AI levels, while TSMC's AI premium remains intact.
What method did HSBC use to reach this conclusion?
HSBC did not forecast profits directly. Instead, it reverse-engineered the earnings path implied by the current share price.
The model starts with three-year consensus earnings, runs 100,000 Monte Carlo simulations — a statistical technique that tests vast numbers of random scenarios to approximate real probabilities — and filters for 15-year earnings trajectories consistent with today's price.
In plain terms = the question is not "how much will Samsung earn?" but "given today's price, what does the market *believe* Samsung will earn over 15 years?" — letting the stock price speak for itself.
How pessimistic is the market on Samsung's earnings?
Since its early-June peak, Samsung's share price has fallen roughly 25%.
The market-implied earnings cycle has shortened from about 3.5 years to about 2.5 years. Implied EPS compound growth from year three to year nine has dropped from roughly −15% to −35% — a historical low.
This means → the market is not merely saying "the AI dividend will fade." It is saying Samsung's earning power after the full cycle will be lower than before AI began — implied trend EPS has collapsed from about 2× to just 0.8× the 2024 level.
How do SK Hynix and TSMC compare?
SK Hynix has adjusted even more sharply: down 37% since its June 25 peak. Its implied earnings cycle crashed from roughly six years to about two years and eight months, and long-term trend earnings fell from about 6× to roughly 2× the 2024 level. HSBC calls this pricing "excessively pessimistic."
TSMC tells a different story: its implied cycle still spans about seven years and four months, with trend earnings at roughly 2.3× the 2024 level.
This reflects a sharp market distinction — the foundry link (TSMC) retains its AI premium; the memory link (the two Korean names) has had it nearly wiped out. This divergence is the defining feature of the current AI-sector correction.
Is the mechanical selling pressure close to ending?
Year-to-date, foreign investors have net sold roughly $150 billion across Samsung, SK Hynix, and TSMC — about $60 billion of that since June alone.
Single-stock 2× leveraged ETFs — funds that use borrowed money to double-down on one stock — are deleveraging fast. Their combined AUM has shrunk from about $37 billion at end-June to $12 billion, and their share of turnover on high-volatility days has dropped markedly.
This means → the most destructive mechanical selling is fading. HSBC argues the next key question is no longer whether AI demand exists, but whether the current price compresses the AI cycle too pessimistically — if demand holds, the long-term earnings assumptions baked into Korean memory stocks may be due for an upward revision.
Content is for reference only, not financial advice.