You've heard the hand-wringing. Taiwan's Taiex and South Korea's Kospi are ripping to records, but it's all just a handful of AI chip stocks. TSMC supposedly weighs over 40% of the Taiex. Samsung and SK Hynix together dominate more than 42% of the Kospi. One wrong move in AI demand, the story goes, and these indexes collapse like a house of cards. The broader market is weak, the rally fake, and concentration risk is about to punish everyone. Classic bubble talk.
Reality is the punchline: this concentration is exactly what a real earnings supercycle looks like. These three companies own the highest-margin, supply-constrained choke points in the entire AI stack—advanced foundry capacity and high-bandwidth memory. Markets aren't distorting anything. They're weighting by economic reality, and that reality is printing money at a scale that lifts entire national market caps past major European economies.
Look at the numbers. TSMC reported Q1 2026 revenue of NT$1.134 trillion, up 35% year-over-year, with net income surging 58% to NT$572.48 billion. Gross margins sat north of 66% in recent prints as AI demand for 3nm and 2nm nodes outstripped everything. The company raised full-year 2026 revenue guidance to over 30% growth in USD terms. That's not hype—it's capacity sold out into next year with pricing power intact.
Samsung's semiconductor division delivered operating profit of KRW 53.7 trillion in Q1 2026—up nearly 50-fold from the year-ago period. That one division accounted for over 90% of the company's record group operating profit of KRW 57.2 trillion. SK Hynix, the HBM leader, posted its own record quarter with revenue nearly tripling year-over-year and operating margins hitting the 70% range on AI memory shortages. These aren't one-off beats. AI capex from the hyperscalers is structural, and these firms control the irreplaceable bits.
The deadpan fact bomb: when two Korean memory makers and one Taiwanese foundry drive the majority of profits and index gains, calling it 'concentration risk' is just complaining that the AI buildout has clear choke points—and these companies own them. Kospi has vaulted toward 7,000-8,000 territory with massive single-day moves led by semis. Taiex has hit repeated records, sitting near 41,000-42,000 levels. Taiwan and Korea's combined market relevance in global equities now punches like economies that dwarf most of Europe. The semis aren't masking weakness; they're the strength.
This is earnings-weighted pricing in real time. Global semiconductor revenue is forecast to exceed $1.3 trillion in 2026, with AI chips making up roughly 30% of that total and memory in tight supply. HBM demand remains insatiable, advanced node utilization near maximum, and ASPs holding or rising. The market isn't lazy for piling into these names—it's correctly ignoring the parts of the economy where economics don't support broad participation. Pretending otherwise is what creates actual bubbles.
You see the same pattern in the U.S. The Magnificent 7 didn't distort the S&P; they earned their weights by delivering the growth. Same dynamic here, just more extreme because the AI stack has even tighter oligopolies at the leading edge. TSMC's foundry dominance and the HBM duopoly aren't going away overnight. Capital spending is ramping, but new supply lags demand by years. That's not risk—it's durable pricing power.
Connect the dots to the macro. AI infrastructure spend isn't optional for big tech. It's table stakes for staying relevant in the next decade. Every hyperscaler update screams more GPUs, more memory, more advanced packaging. That flows straight to these choke-point players. The rest of the index gets a halo effect—supply chains, adjacent tech, even local economies—but the profits are overwhelmingly concentrated where the value is created. That's healthy, not distorted.
Of course, nothing lasts forever without checks. The kill criteria are clear and measurable: if TSMC, Samsung, and SK Hynix combined miss 2026 revenue growth guidance by more than 10% or slash aggregate capex below $50 billion in the next two quarters, the thesis weakens fast. HBM and leading-edge utilization dropping below 85% or ASPs falling over 15% quarter-over-quarter by Q3 2026 would signal real demand cracks. And if the Kospi and Taiex ex the top names show flat or negative earnings growth while the big three slow, that's your cue the broader story was always fragile.
Until then, stop fearing the concentration. Embrace it. These aren't just stocks driving indexes—they're the companies turning AI from PowerPoint slides into actual economic output at massive margins. The rally in Taiex and Kospi isn't a warning sign. It's the scoreboard.
The market got this one right. The real risk is being the guy still waiting for 'breadth' in a winner-take-most cycle.