Stocks for Thoughts

Thoughts on the Korea Trade

You watched the US spend on AI. Korea is the country that gets paid when they do — and its stock market is a more concentrated AI bet than the S&P itself. Two companies carrying a whole country. I don't decide the ending before we look. Open any box below.

Margin the Fox
Living log · newest entry open · figures 2026, re-verify before relying
Margin the Fox, Korea edition — a fox in a suit (placeholder until a Seoul illustration is made)
How to read this log: each box follows the money on one piece of the Korea trade and lands where the evidence lands. The map stays first; the newest entry opens by default. Dashed boxes are placeholders for analysis still being built. Figures are 2026 reporting and should be re-verified before relying on them. A framework for thinking, not a forecast, recommendation, or investment advice.
1 The map · concentration Two stocks. Almost half the index. Foundational map
Margin the Fox — laying out the map

Forget "Korea is booming" for a second. The real question is what is doing the booming. The US AI trade is roughly 45% of the S&P 500 — but diffused across the Magnificent Seven and beyond. Korea's version is far more extreme: not a theme spread across dozens of names, but two companies carrying a whole index.

~24%Samsung
~18%SK Hynix
~58%the other ~950 companies
Samsung + SK Hynix = a record 42.2% of the entire KOSPI (May 2026)

Two stocks at a higher share of one index than the entire AI theme is of the S&P 500. Buy "Korea," and you've mostly bought memory chips.

Market cap past $1 trillion
Samsung Electronics
~+330%
stock move · trailing 12 months
The HBM standout
SK Hynix
~+800%
stock move · trailing 12 months
~+95%
The KOSPI's gain year-to-date in 2026, hitting fresh records — powered almost entirely by memory chips. The question isn't "is Korea booming." It's "what exactly is doing the booming."

Sources: KOSPI ~+95% YTD 2026 (record highs; reports ranged +95–100%). Samsung + SK Hynix = record 42.2% of the KOSPI (May 2026, Manulife). US AI theme ~45% of S&P 500 but diffused. Samsung ~+330% / SK Hynix ~+800% trailing 12mo (figure varies by data vendor/date; SK Hynix's widely-cited ~+300% is the 2026 YTD); both market cap past $1T. Figures 2026 reporting — re-verify. Not investment advice.

2 Why it isn't hype · the chokepoint The one chip Nvidia can't get enough of The moat, measured
Margin the Fox — the moat is real

Here's what separates Korea from a generic "chip rally." SK Hynix is the dominant supplier of HBM — the high-bandwidth memory that sits directly on Nvidia's H100 and H200 chips and makes them work. And it's structurally scarce: harder to make, lower yields, lead times stretching past a year. Nvidia has publicly said it can't get enough.

That scarcity is real pricing power, not a story. The two Korean giants control about two-thirds of the world's DRAM (~67%) and have pushed for server-memory price hikes of up to 70%. This is a supply-constrained duopoly printing money — very different from a bubble built on hope.

~67%
global DRAM share — the duopoly
+151%
March chip exports, year-over-year
~60%
of Korea's revenue is AI-linked (Goldman)

Sources: SK Hynix dominant HBM supplier (H100/H200; ~year-plus lead times; Nvidia supply-starved; SK Hynix ~50–62% of HBM). Two firms ~67% global DRAM (Samsung 38.6% + SK Hynix 28.8%, Q1 2026 Omdia); sought 60–70% server-DRAM price hikes (Q1 2026). March 2026 semiconductor exports +151% YoY (topped $30bn). Goldman (Tim Moe): Korea ~60% AI-revenue-exposed. Figures 2026 reporting — re-verify. Not investment advice.

3 Newest · the tell The central bank is watching the chip guys' credit cards. Jun 2026
Margin the Fox — concentration cuts both ways

Want proof this concentration is real money and not a chart story? On June 17 the Bank of Korea named two companies as an inflation risk — not a sector, two companies. Samsung and SK Hynix bonuses, up to ~$455K per worker, are now a variable in the central bank's price outlook.

The numbers are absurd: a memory worker on a ~$52K base could pocket a ~$410K bonus this year. IT-sector special pay jumped 60.6% year-over-year last quarter while everyone else got 2.1% — a gap so wide that labour groups now cite the chip bonuses as a benchmark in minimum-wage talks. The money is leaking out of two companies and into the whole economy's wages.

And you can see it on the receipts.

// Gyeonggi Province · the semiconductor belt
Luxury jewellery sales (Shinsegae, near the fabs)+146% YoY
Luxury watch sales+85% YoY
Shinsegae (dept-store operator) — stock+190% YTD
Lotte Shopping — stock+148% YTD
Hyundai Department Store — stock+119% YTD

The market followed the money downstream — re-rating the department stores as chip-boom plays. The boom got so big it re-rated the malls where the chip guys shop.

Two stocks 42% of the index a few zip codes the national inflation rate.

Margin's read: The KOSPI is a memory cycle wearing a country's clothes. Index-level gains say less about "broad Korean strength" than about the earnings power of a narrow group of exporters. When you buy the KOSPI, you're buying the most leveraged, concentrated expression of the global AI-spending cycle anywhere on Earth — more concentrated than the S&P. That's the strength and the whole risk: the pricing power is genuine and the export boom is real, but the same concentration means whatever moves AI capex moves Korea, hard, in both directions.

One honest caveat: the BOK itself says this year's inflation has been largely energy-driven (the Iran war), with the bonuses the emerging risk rather than today's main driver. The point isn't that chip guys broke the CPI. It's that two companies now move the needle at all.

Sources: BOK price-stability report (June 17, 2026): IT special pay +60.6% YoY vs 2.1% elsewhere; bonuses up to ~$455K/worker (~$52K base + ~$410K bonus cited); Gyeonggi luxury +146% jewellery / +85% watches; Shinsegae +190% / Lotte +148% / Hyundai Dept Store +119% YTD. BOK flags inflation as largely energy-driven (Iran war), bonuses an emerging risk. Figures 2026 reporting on a fast-moving situation — re-verify before relying. Not investment advice.

4 Placeholder · next post Peak-cycle, or a new structural base? Durability — the question that decides it
Margin the Fox — eyeing the durability question

The pricing power is real and the export boom is real — but is this earnings boom peak-cycle or a new structural base? That's the question that actually decides durability, and it's the one we pull apart next. The strange tell to start from: the stock tripled and got cheaper — earnings ran faster than the price. Whether that re-rating is the market seeing a durable HBM franchise, or just front-running a memory cycle that always mean-reverts, stands up as its own entry.