Title: 🧠 SK Hynix Bets AI Broke the Cycle — The AI memory leader is coming to Nasdaq Show: App Economy Insights / How They Make Money (Substack) — Premium Guest: Bertrand (App Economy Insights) Date: 2026-JUL-07 URL: https://www.appeconomyinsights.com/p/sk-hynix-bets-ai-broke-the-cycle Length: written post (no timestamps) Note: SK Hynix's ~$28B Nasdaq ADR listing (SKHY, ~Jul 10): HBM leadership (~57% share, ~2/3 of NVIDIA HBM4), record Q1 FY26 (revenue +198% Y/Y, 72% op margin), priced at parity with Micron ~7x forward; author's take = watchlist, don't pay peak margins as a new baseline. Body reproduced for personal study; promo/subscription boilerplate omitted.
SK Hynix is trying to pull off the largest ADR listing in history.
The planned Nasdaq listing targets over $28 billion (45 trillion won), edging past Alibaba's ~$22 billion New York debut in 2014. Trading is expected to begin around July 10. Every dollar is earmarked for fabs, packaging, and lithography machines. The funny part is that the company doesn't really need the money.
SK Hynix is the world's #1 HBM maker (High Bandwidth Memory) at ~57% share. The company sold out of capacity three years in advance, and it just surpassed Samsung to become South Korea's most valuable company. It also already trades right in line with Micron, near 7x forward earnings. The re-rating a Nasdaq listing is supposed to deliver already happened.
We broke down Micron's blowout quarter last week, noting that the company benefited from the AI memory shortage.
This week, the leader that ships the majority of NVIDIA's HBM is asking US investors to price it like the AI infrastructure play it has become.
Today at a glance: 📈 The US Listing · 🧠 How SK Hynix Makes Money · 📊 Q1 FY26 in Numbers · 🇰🇷 What You're Paying · ⚔️ Competition & Risks · 🔭 What to Watch · 🧭 Personal Take
📈 The US Listing
SK Hynix shares trade in Seoul, in won, on the Korea Exchange. That makes them awkward and expensive for US institutions to touch directly. An American Depositary Receipt (ADR) solves that. The company issues new shares, parks them with a custodian bank, and a US depositary issues receipts that trade on Nasdaq in dollars like any American stock.
SK Hynix plans to issue up to 178 million ADRs, the equivalent of 2.5% of the company. Each Seoul share trades near $1,580 (₩2,425,000) and splits into ten ADRs, so a single ADR is referenced at around $158 (₩242,500), with the final price set on July 10. The ADRs list on the Nasdaq Global Select Market under SKHY.
Every dollar is going into capacity.
The proceeds are split four ways:
🏭 Yongin Y1: the first fab at SK Hynix's sprawling new Yongin cluster, its next major DRAM and HBM base.
📦 Cheongju P&T7: an advanced packaging plant built for HBM, where the die-stacking that sets SK Hynix apart actually happens.
🔬 ASML EUV scanners: the extreme-ultraviolet lithography tools that leading-edge DRAM can't be made without.
🇺🇸 Indiana: a $4 billion packaging plant, SK Hynix's first US fab.
Compare that to the SpaceX IPO. SpaceX is raising funds for its CapEx ramp, with $10 billion in negative free cash flow in Q1 alone. SK Hynix is raising money from a position of strength, with $24 billion in net cash and an order book sold out through 2028.
SK Hynix first floated a raise of around $10 billion, and the board eventually settled on $28 billion. You don't do that unless internal demand forecasts have moved well past the old memory-cycle playbook.
There's a control wrinkle worth mentioning. SK Square, the holding company that owns ~20% of SK Hynix, must keep its stake above 20% under Korea's holding-company rules. That constraint is why the deal issues new shares sized to protect that floor, rather than selling treasury stock. Some Korean shareholders are unhappy about the dilution. For US investors, the dilution is the price of admission to a stock they couldn't easily buy before.
🧠 How SK Hynix Makes Money
The comeback
SK Hynix started in 1983 as Hyundai Electronics, the chip arm of the Korean industrial group. The Asian financial crisis reshaped it. Seoul forced a 1999 merger with rival LG Semicon, creating a top-tier DRAM maker overnight and burying it in debt. When memory prices collapsed ~80% in 2001, the renamed Hynix lost billions, fell under creditor control, and became a national symbol of corporate failure.
Rescue came in stages. In 2002, Micron offered ~$3.5 billion for Hynix's memory business, and the board rejected it to go it alone under $6.5 billion of debt. SK Group bought control in 2012, renamed the company SK Hynix, and gave a serial cycle-survivor its first patient parent. That stability funded an early bet on HBM, a niche product for years before AI made it the most valuable memory on earth.
Three products, one bottleneck
SK Hynix is a memory company. It makes two families of chips:
DRAM is the fast, volatile memory that feeds processors.
NAND flash is the slower, persistent storage inside SSDs. A small foundry arm and Solidigm, its enterprise-SSD subsidiary, fill out the rest.
According to IDC, SK Hynix has the following revenue market share for Q1 2026:
#2 in DRAM at 29.1% (Samsung reclaimed the top spot in late 2025).
#2 in NAND at 18.5%.
#1 in HBM at 56.4%. The last one is the whole story.
HBM, or high-bandwidth memory, is DRAM stacked vertically and bonded into a tower that sits right beside an AI GPU. As AI models grew, the constraint shifted from raw compute to how fast you could feed the chip. HBM is the answer, and SK Hynix has been first to market on every generation since HBM2E.
NVIDIA buys roughly two-thirds of its HBM4 from SK Hynix for the Vera Rubin platform. The edge in HBM goes well beyond process technology. It comes from TSV (through-silicon vias), packaging, yield, and a track record of never burning the customer. HBM climbed to 12% of DRAM revenue in Q1, double the 6% a year earlier, and it carries the richest margins in the business.
Why it matters
Business model: Memory has always been a commodity sold by the bit. HBM flips that, at least for now. It's a custom, co-engineered product tied to a customer's chip roadmap, which makes it stickier and much harder to discount.
Competitive moat: Being NVIDIA's lead HBM supplier means SK Hynix sits inside the most important hardware roadmap in tech. Displacing it requires matching yield and timing on a moving target.
Investor angle: If HBM remains a critical part of the SK Hynix franchise, the company could see its multiple re-rate. If it becomes a three-way commodity, it stays ordinary memory.
Takeaway: SK Hynix doesn't need to win all of memory. It needs to keep owning HBM, the one slice that AI can't function without and rivals can't yet replicate at scale.
📊 Q1 FY26 in Numbers
Q1 FY26 (March quarter) was the strongest in the company's history by every metric.
Income statement:
Revenue grew +198% Y/Y to $35.5 billion (52.6 trillion won). That's the first quarter ever above 50 trillion won, and more than SK Hynix booked in all of 2023.
🧠 DRAM (including HBM): the majority of revenue, with DRAM ASP (Average Selling Price) up mid-60% Q/Q.
💾 NAND flash: ASP rose sharply, with shipments down ~10% Q/Q as the mix moved toward high-value products.
Operating margin was 72% (+13pp Q/Q). That's an all-time high, above NVIDIA's or TSMC's most recent prints. Operating profit was $25.4 billion (37.6 trillion won, +405% Y/Y).
Net margin was 77%. Net profit was $27.3 billion (40.3 trillion won, +398% Y/Y).
Balance sheet:
Cash and equivalents: ~$37 billion (54.3 trillion won), up 19.4 trillion won in a single quarter.
Net cash: ~$24 billion (35 trillion won), after debt fell to 19.3 trillion won.
Guidance:
For full-year context: FY25 was already a record. Revenue rose +47% Y/Y to 97.1 trillion won (~$66 billion), with operating profit +101% and net profit +117%. Then Q1 FY26 alone produced more operating profit than all of FY24. At Q1's run rate, FY26 revenue would more than double last year's.
So, what to make of all this?
📈 Price is the key. Like Micron last week, almost none of the growth came from shipping more bits. DRAM ASP jumped mid-60% in one quarter against roughly flat shipments. The shortage is showing up in prices, and SK Hynix is best-positioned to capture it.
🧠 HBM is the margin engine. Doubling HBM's share of DRAM revenue to 12% is why a memory company is printing a 72% operating margin. Strip HBM out, and these would be good-but-ordinary numbers. With it, the P&L looks NVIDIA-esque.
🌊 Seasonality didn't matter. Q1 is normally the weak quarter for memory. It set records anyway, because AI capex doesn't follow the consumer-electronics calendar.
🔒 Demand exceeds three years of capacity. Management says HBM orders already outrun planned output through 2028. It's turning these cyclical stocks into a "structural demand" story.
💰 The cash flips the model. A memory maker with $24 billion of net cash and a 79% EBITDA margin is a different animal than the debt-laden SK Hynix of past downturns. The balance sheet has turned into a weapon.
⚠️ Margins are already peak-like. A 72% operating margin is nowhere near normal for memory, even if the cycle itself has room to run. The question every buyer has to answer is not whether HBM demand stays strong next quarter. It's how much of this margin profile survives when supply catches up.
Key takeaway: SK Hynix is having the best quarter of its life, and the top may still be many quarters away. That's exactly what makes the stock hard. The bull case is that HBM and long-term agreements have structurally raised the floor. The bear case is that memory always reverts, and that 72% margin is someone else's opportunity.
🇰🇷 What You're Paying
SK Hynix comes to market priced almost exactly like Micron. The so-called "Korea discount" for weak governance and limited access is gone. Both trade near ~7x forward earnings and ~18x trailing EV/EBIT. The market is treating the richest margins in memory history as close to a new baseline.
Memory has always been cyclical, and margins like these have normalized in every prior cycle. The bull case is that AI changed the pattern. HBM is sold out through 2028, priced under multi-year contracts, and tied to demand that looks structural rather than seasonal. The bear case is that memory is memory, and every past boom claimed to be different, too.
The ADR's real function is access. US institutions can finally own the HBM leader directly, in dollars, with the liquidity and index eligibility a Nasdaq line brings. That changes who can buy, but doesn't make the stock cheap.
Why it matters
Business model: The reframing from "cyclical memory producer" to "AI infrastructure company" is already priced in. From here, the multiple holds only if the earnings do.
Access and flows: A Nasdaq line widens the buyer base to US funds and index trackers that couldn't touch the Seoul shares. That supports the price through demand, separate from fundamentals.
Investor angle: At parity with Micron, SK Hynix requires making the same peak-cycle bet on the same memory boom. The only difference is that the company happens to lead HBM.
Takeaway: The discount closed before the bell. What the listing sells is access, liquidity, and index inclusion. The one open question is whether the leader eventually earns a premium over Micron, and the market isn't paying it yet.
⚔️ Competition & Risks
SK Hynix may be the cleanest AI-memory story on the board. That doesn't make it low-risk. A 56% HBM share looks unassailable, but the lead is newer and more contestable than that number suggests.
⚔️ Samsung is back: Samsung has passed NVIDIA's HBM4 qualification and begun shipments. Counterpoint projects that it will capture ~28% of the 2026 HBM4 market, compared with SK Hynix's ~54%. The near-monopoly SK Hynix enjoyed on early HBM4 supply could shrink from roughly six months to a single quarter. SK Hynix still leads. But don't count Samsung out.
🇨🇳 China at the low end: China's memory champions, CXMT in DRAM and Yangtze Memory in NAND, are scaling fast with heavy state backing. They trail badly at the high end, and neither is a real HBM threat in 2026. But standard DRAM is fungible, and volume is how that gap could eventually close. The real pressure lands in 2027 and later, once that volume shows up.
🎯 Customer concentration: A franchise built on ~two-thirds of NVIDIA's HBM4 is also an exposure to one company's roadmap, pricing leverage, and appetite for second sources. NVIDIA is already building those second sources by pulling Samsung and Micron into the mix.
🌊 The cycle, always the cycle: Memory's history is overbuild, glut, crater, repeat. Every prior peak in this industry looked permanent right up until it wasn't. Sold-out capacity and multi-year agreements are real but unproven across a full down-cycle.
💧 Dilution and FX: The offering itself adds ~2.5% to the share count, and US investors take on won/dollar exposure plus the governance and geopolitical factors that created the discount. A re-rating would close the gap. It wouldn't erase the risks behind it.
Takeaway: The bear case has nothing to do with SK Hynix being weak. The two pillars under the valuation, HBM dominance and peak-like margins, are exactly what competitors and the cycle are built to erode. The listing prices assume it can hold its lead.
🔭 What to Watch
🤖 The HBM4 split with Samsung. The single most important number next year is what share of NVIDIA's Vera Rubin HBM4 SK Hynix will actually win versus Samsung. Watch for formal allocation and pricing. A narrowing split would show a narrower moat in real time.
💹 The ADR debut is a signal. Pricing and the first weeks of trading are the most informative read available. A strong open that closes the Micron gap would validate the re-rating thesis. A weak one would show that US investors still see a Korean memory cyclical, US ticker or not.
🏗️ CapEx discipline and Yongin timing. SK Hynix is spending hard: Yongin's Y1 fab pulled forward to early 2027, M15X ramping, and billions in EUV tools on order. The test is whether it can add capacity fast enough to hold the lead without flooding a market it wants to keep tight.
📉 The pricing curve into 2027. Watch DRAM and NAND ASPs quarter to quarter. The bull thesis doesn't require prices to keep spiking. It requires them to stay elevated as new supply arrives. The first quarter of decelerating ASPs will be when the "structural vs. cyclical" debate gets settled.
🧭 Personal Take
My golden rule for IPOs is that IPO stands for "It's Probably Overpriced."
This one comes with a twist, because SK Hynix is already public and already wildly profitable, and it's raising only to build. Survival was never the risk here. The real question is the price you pay for peak-cycle earnings, and a market that keeps getting more crowded.
SK Hynix is the leader on every axis that counts: best product, lowest cost, highest margin. Now it wants to be valued like one.
📈 Bull Case
The HBM leader is sold out through 2028, sitting on $24 billion of net cash, with a margin profile that would embarrass most of the Mag 7. Now it's finally accessible to the deepest pool of AI capital on earth. If the long-term agreements hold through a softer patch, the multiple can hold too.
📉 Bear Case
You're buying peak margins at peak narrative. Samsung is back in HBM4, China is grinding up from the bottom, and the entire bull case rests on the one thing memory has never delivered: a cycle that doesn't turn. Pay 7x forward earnings for trough-of-the-cycle SK Hynix, and it's a steal. Pay it for peak-of-the-cycle SK Hynix, and the math looks very different. Earnings multiples don't say much without cycle context.
Takeaway: SK Hynix belongs on every serious investor's watchlist. It is the cleanest public expression of the AI memory bottleneck. But the listing arrives at record margins treated as a new baseline. The market has already embraced the story. I'd rather watch the ADR trade through a quarter or two than pay up for permanence the industry has never delivered.
Disclosure (author): I own AMD, AMZN, ASML, AVGO, GOOG, META, NVDA, TSLA, and TSM in App Economy Portfolio.