The Kospi tripled in 18 months. The Taiex is up 44% in 2026. Both rallies are led by essentially the same story: a shortage of AI memory and logic chips sending a handful of names parabolic.
That's not a stock picking problem. That's a benchmark problem - because when three stocks control half an index, the word "diversification" stops meaning anything and the reference point for every relative judgment collapses.
The question isn't whether these stocks are overvalued or undervalued in absolute terms. It's whether the concentration itself has rewired what "beating the market" even means.
Taiwan: TSMC and the 42% Index
TSMC now represents roughly 42% of the Taiex benchmark and 58.33% of the MSCI Taiwan Index after MSCI raised Taiwan's weighting in May. That is up from roughly 15% a decade ago. The Taiex has climbed 44% in 2026, and the index fund flows required to rebalance that upward weighting create a self-reinforcing buying cycle.
In our book, you judge every stock against its sector peer set. When one name is 58% of the index, the peer set is 42 names sharing the remaining 42%. That's not a comparison set - that's a rounding error.
The mechanical consequence showed up in April. Taiwan's Financial Supervisory Commission raised the single-stock cap for active domestic funds from 10% to 18%, because managers were stuck: underweight TSMC and lag the benchmark, overweight it and break concentration rules. Funds worth roughly $100 billion in Taiwan were effectively forced to sit on the sidelines of their own market. The rule change sent TSMC to a record high the next day as newly unlockable capital flooded in.
That's not a commentary on TSMC's quality. It's a commentary on what happens when benchmark construction and concentration rules collide. Active managers didn't need a thesis - they needed permission.
What the factor stack says: If your process requires a comparison set and the comparison set has been reduced to a minority fragment of the index, the process either adapts or it produces garbage. In a 58%-weighted index, the only relative judgment that matters is TSMC vs. the rest. The rest gets the same grade by default because nothing in the rest moves the needle.
South Korea: Samsung, SK Hynix, and the Margin-Fueled 105%
South Korea is running a parallel playbook, but with more leverage in the room.
Samsung Electronics and SK Hynix together account for roughly 42% of the KOSPI - Samsung at 27%, SK Hynix at 25%. The KOSPI has surged 105% over 18 months, with SK Hynix alone up 265% since November. SK Hynix joined the $1 trillion market-cap club in late May, sitting at roughly ₩1.39 quadrillion in market value.
Here's the part that isn't in the headlines yet. Korean retail investors are financing this with record levels of margin debt. Government officials have publicly warned about the leverage. Then on June 4–5, foreigners sold ₩3.5 trillion ($2.5 billion) in a single session and the Kospi fell 5.5%, with Samsung and SK Hynix each sinking more than 6%. The rally's two pillars sold off in unison and there was nobody else in the index to cushion the drop.
The FT called the KOSPI rally one that outpaced Nasdaq's dotcom-era gains. That's a useful comparison only if you remember what happened to breadth after the dotcom index peaked.
What the factor stack says: Momentum and revisions are timing tools, not standalone theses. SK Hynix up 265% is impressive momentum. But momentum on a 42%-weighted pair in a margin-fueled market is a different animal than momentum in a normally distributed index. The signal is still there, but the noise is structural - the index can't decouple from its two largest names, and when they wobble, the whole thing wobbles.
What This Means for How You Actually Invest
Neither Taiwan nor South Korea is breaking because of the underlying businesses. TSMC controls the advanced-node foundry market and its margin structure remains structural, not cyclical. Samsung and SK Hynix are riding the HBM (high-bandwidth memory) cycle, which is the memory technology inside every major AI accelerator, and demand is still outstripping supply.
The problem isn't the thesis. It's the plumbing.
When two names dominate an index, three things happen:
Active managers underperform mechanically. You can't diversify away from a 42% position without accepting persistent lag. The underperformance isn't a skill problem - it's a math problem.
Passive flows amplify the cycle. Every dollar of new inflow into a broad Asian index buys the same concentrated basket. The concentration begets more concentration.
Relative valuation loses its anchor. When you screen for "cheap" stocks in Korea and find one trading at half the sector median PE, that cheapness is partly real and partly a function of everything outside Samsung and SK Hynix being crowded out of capital. A D-grade valuation can look like a C+ if the only alternative is a 42% weighting in two stocks you can't differentiate.
So what do you do when the process is technically sound but the comparison set is broken?
You adjust the comparison set. Instead of "cheap vs. KOSPI," you look at "cheap vs. global semiconductor peers" - because the relevant universe for Samsung's valuation isn't the KOSPI, it's the global chip sector. TSMC's comparison isn't the Taiex, it's ASML, Applied Materials, and GlobalFoundries. The index weighting is a domestic plumbing issue, not a fundamental truth about where the stock sits.
That's the distinction. The indices are warped. The factor stack doesn't have to be - as long as you refuse to let a benchmark's mechanical concentration substitute for a real comparison.
The risk flag to watch: Korea's margin-fueled breadth. If Samsung and SK Hynix pull back another 10-15% on foreign selling and retail deleveraging, the KOSPI drops with them because there's nothing else in the index to hold the line. In a normally distributed market, that wouldn't happen. In a 42%-concentrated market, it's the baseline scenario.
TSMC in Taiwan faces the same concentration mechanics but with less retail leverage hanging over it. The April rule change was the government admitting the system had broken. It hasn't been fully fixed - just loosened enough to keep flowing.
Volatility in these markets isn't a signal that the thesis is wrong. It's a signal that the concentration is real, the leverage is real, and the only hedge against a concentrated rally collapsing is to not be concentrated in the first place.