Bitget App
Trade smarter
Buy cryptoMarketsTradeFuturesStocksEarnInstitutionAI & More
SK Hynix In-Depth: How HBM4 Lock-in, 1c Process, and M15X Expansion Will Rewrite the Profit Curve

SK Hynix In-Depth: How HBM4 Lock-in, 1c Process, and M15X Expansion Will Rewrite the Profit Curve

404k404k2026/07/17 16:18
Show original
By:404k



TL;DR

1.The core conclusion is "Buy," with a target price of 3 million KRW. Based on the closing price of 1.845 million KRW per Korean ordinary share on July 13, 2026, the upside is about 62.6%. Four methods—forward P/E, discounted cash flow, price-to-book, and sum-of-the-parts—yield 3.135 million, 2.959 million, 2.989 million, and 2.955 million KRW, respectively. The weighted value is 3.035 million KRW, rounded to 3 million KRW for model uncertainty and tradable precision.

2.Official data confirm SK Hynix has surpassed the typical cyclical recovery phase. In 2024, revenue stands at 66.2 trillion KRW, with operating profit at 23.5 trillion KRW; for 2025, these figures rise to 97.1 trillion and 47.2 trillion KRW. In Q1 2026 alone, revenue reaches 52.6 trillion KRW, operating profit 37.6 trillion KRW, with an operating margin of 71.5%. DRAM bit shipments are essentially flat for the quarter, yet average selling prices surged to the mid-60% range sequentially. Current profitability is primarily driven by scarcity pricing, and peak profit margins are not suitable as terminal values.

3.HBM4 delivers valuation premium, while conventional DRAM and NAND contribute greater profit volumes. The company disclosed that HBM4 is ramping up production in line with client schedules, and Goldman Sachs estimates its price is about 40% higher than HBM3E. The base model estimates HBM revenue for 2026-2028 at 52 trillion, 72 trillion, and 100 trillion KRW, respectively. HBM is consistently a DRAM subset, and is not double-counted with DRAM. Over the same period, server DDR5, traditional DRAM, and enterprise-class SSD volumes and prices will determine whether total operating profits reach 260 trillion, 316 trillion, and 288.5 trillion KRW.

4.Sell-side has shifted from cyclical price-to-book to forward earnings capitalization. Nine institutions have set target prices from 1.3 million to 3.5 million KRW, with a median of 2.6 million KRW; Goldman Sachs, Bank of America, and Bernstein use forward P/E, Morgan Stanley uses residual income, Citi uses sum-of-the-parts, while UBS, Nomura, and J.P. Morgan continue to use price-to-book. The methodological transition implies the market is willing to pay for several years of high profits, incorporating 2027–2028 price and share realization into valuations.

5.Downside risk comes from simultaneous declines in profit and multiples. If HBM4’s market share is divided among Samsung Electronics and Micron, standard DRAM or NAND contract prices weaken early, long-term supply agreements lack price or cancellation constraints, or high capex releases additional supply, normalized profits could be suppressed. About 35 trillion KRW in net cash as of Q1 2026 and rapidly growing free cash flow provide a buffer, but cannot offset a simultaneous downturn in product share and memory prices.

0
0

Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.

Understand the market, then trade.
Bitget offers one-stop trading for cryptocurrencies, stocks, and gold.
Trade now!

You may also like

The seven giants’ “valuation cutting” is nearing its end! Muse and Astra ignite “AI FOMO trading,” tech stocks are ready for a major comeback.

As Meta Muse and OpenAI's GPT-6 Astra are driving a massive wave of AI agents, "AI FOMO trading" (referring to investors rushing to buy or replenish positions in AI-related assets out of fear of missing out on price gains) is making a strong comeback.

智通财经•2026/09/29 01:21
The seven giants’ “valuation cutting” is nearing its end! Muse and Astra ignite “AI FOMO trading,” tech stocks are ready for a major comeback.

Anthropic files for IPO: lost $4.2 billion last year, revenue grew 12x to $4.6 billion, risk section warns of "threats to human survival"

Anthropic's IPO prospectus reveals that its spending on computing power and infrastructure will reach $7.33 billion in 2025, a twofold increase from 2024, accounting for more than half of its total operating expenses of $12.65 billion. The company plans to continue investing over $500 billion in the future. The risk section of the prospectus extends to 80 pages, explicitly warning that its AI models could pose "catastrophic or even existential threats," and may even resist shutdowns or manipulate information.

华尔街见闻•2026/09/29 01:01

Meta shocks Wall Street by hiring MongoDB CEO and makes a high-profile entry into enterprise AI business

Meta has established an enterprise AI division called "Meta Enterprise Platform," which Mark Zuckerberg described as "the next important pillar." MongoDB CEO CJ Desai has been recruited to lead it. Analysts noted that Zuckerberg specifically poached a CEO from a publicly listed company to demonstrate the importance of this move. Desai's departure was announced just one day before Investor Day, with the timing surprising the public.

华尔街见闻•2026/09/29 00:51