Three Memory Titans: The Leader, The Integrator, and the Only Pure Play on US Markets

「Industry Research」Memory Series, Part 3 — and the finale of the memory coverage. The first two pieces covered why this cycle is different and why HBM is the bottleneck. This one gets to the most concrete question: among the three giants, who is who, and who is more likely to come out on top?

011. Three Different Playbooks

Many people lump SK Hynix, Samsung, and Micron together as "three memory makers," thinking they are just three players in the same business. That is a crude misconception.

The truth is — in this supercycle, the three companies are playing completely different games. Their advantages, risks, and investment logic are almost three separate stories:

• SK Hynix: The leader that has bet everything on HBM. The purest and sharpest, but also the most dependent on a single narrative.
• Samsung: The only integrator with the full "memory + logic foundry + advanced packaging" stack. The deepest pockets, but the elephant turns slowly, and it has actually fallen behind in HBM.
• Micron (MU): The only pure play directly accessible on US markets, and riding the tailwind of US semiconductor policy. High beta, but it is still a laggard with cyclical characteristics.

Let's break them down one by one. By the end, you'll realize that "buying memory" is far from "betting on which stock will rise the most" — you first need to figure out which story you are buying.

022. SK Hynix: The Leader Betting Everything on HBM

Among the three, SK Hynix is the sharpest and purest knife of this cycle.

Its core advantage is its first-mover position in HBM: roughly 62% HBM market share, world-first completion of HBM4 development (+40% energy efficiency, 10 Gbps), deep ties with Nvidia, and full-year HBM capacity already pre-booked by customers. This lead has produced a nearly surreal financial report — gross margin 79%, operating margin 72%, repeatedly hitting record highs, with a quarterly operating margin surpassing TSMC.

Its strategy can be summed up in one sentence — Focus, Lead, Lock in with Long-Term Contracts. Concentrate resources on the highest-value segment of HBM, establish a position through technology first-mover advantage, then monetize that position into predictable cash flow through multi-year contracts. This is a very "sharp" approach, and as long as HBM is booming, the returns are the richest.

But the flip side of "sharp" is "fragile." SK Hynix's fate is highly and systematically tied to the single narrative of HBM. It has no logic foundry business like Samsung to cushion the blow, nor the domestic policy protection of Micron. If HBM sentiment turns due to a slowdown in AI capex, it will have the most upside — but also be the first to feel the pain. Its high returns are essentially the reward for "concentrated betting on the single strongest track" — and concentration is always a double-edged sword (remember what I repeatedly said in the piece on Munger and the piece on pseudo-diversification).

My take on it: The purest HBM bullish vehicle of this cycle — highest upside when it works, but its risk is the least diversified.

033. Samsung: The Only "Full-Stack" Integrator, But the Elephant Turns Slowly

Samsung is the hardest to evaluate among the three, because its strength and weakness come from the same thing — size and full-stack capability.

Samsung is the overall global memory leader (roughly 38% share in general DRAM, vs. Hynix's 29%), and it is the only player among the three that possesses the full "memory + logic foundry + advanced packaging" stack. In theory, this is an unbeatable configuration: the HBM4 generation requires "top-tier memory + top-tier logic foundry + top-tier packaging" to work together, right? Samsung is the only one that has all three in-house.

But in reality, on the most critical battlefield of HBM, Samsung is the one falling behind — with an HBM share of about 17%, now overtaken by Micron, and HBM4 once delayed due to yield issues.

This is a fascinating paradox: the giant with the deepest pockets and theoretically the most complete configuration has fallen behind in a new frontier. Why?

My observation: the elephant turns slowly. Samsung's size, its sprawling businesses spanning phones/displays/memory/foundry, and its organizational inertia make it far less decisive than SK Hynix in going "all-in" on HBM. While SK Hynix was pouring all resources into HBM, Samsung was fighting on multiple fronts and balancing trade-offs. The result: in a window where technology iterates fast and first-movers can dominate, the largest player, because it turned slowly, ceded the advantage to a more focused competitor. This is almost a script that business history repeats (remember Christensen's The Innovator's Dilemma — leading incumbents often lose by being "not fast enough to embrace the new thing that will disrupt them").

But by no means should you write off Samsung. Its full-stack capability is a real moat. Once HBM4 yields break through, its catch-up could be rapid (memory market share can shift within one or two quarters). And its full stack gives it a theoretical synergy advantage that no one else has in the trend of "memory + logic" fusion in HBM4.

My take on it: The most underestimated and the most patience-testing one — its story is "whether the integrator's counterattack can succeed," and the bet is on whether the elephant will eventually turn around.

044. Micron: The Only Pure Play on US Markets, Riding the Policy Tailwind

For this site's readers — US equity investors — Micron (MU) is the most practical of the three, for a simple reason: Hynix and Samsung are Korean stocks; Micron is the only pure memory play directly investable on US exchanges.

But Micron's value is far more than just "accessible in US markets." Its fundamentals this cycle are genuinely strong:

Micron's approach sits between Hynix and Samsung — it does not have Hynix's HBM leadership position, but it is rapidly closing the gap (overtaking Samsung in share is a signal); it does not have Samsung's full stack, but it is focused enough. It is a "fast catch-up, and doing it with style" player.

And what truly makes Micron unique — something Hynix and Samsung lack — is the policy tailwind. In the context of US-China semiconductor rivalry and America's push for chip localization (CHIPS Act, domestic fabs), Micron, as the US-based memory leader, is one of the most direct beneficiaries of this industrial policy. The US has mentioned imposing high tariffs on semiconductors not manufactured in America. Such a policy environment is a structural positive for Micron, which has large-scale production in the US, and a source of uncertainty for Korean players. In an era where geopolitics are reshaping supply chains, Micron's "American identity" itself is an asset.

My take on it: The most direct vehicle for US investors to bet on the memory supercycle — strong fundamentals, policy tailwind — but be clear-eyed: it is still the catch-up player among the three, and as the "purest" memory stock, its cyclical volatility is also the greatest. It surges hard, and it can plunge just as hard when corrections come (last year's massive price swings already demonstrate this).

055. Who Will Come Out on Top? My Framework

Now it's time to answer that question: among the three, who is more likely to come out on top?

My honest answer is: that question is wrongly framed. Because "come out on top" assumes a zero-sum game with a single winner. But the HBM track is likely an oligopoly where all three can do well, but with different roles and positions. The real question is not "who wins," but "under different scenarios, whose odds are best?"

Let me use a simple framework to position (not predict, but position):

If you bet on "HBM prosperity continues, technology leader takes all" — then SK Hynix's purity and first-mover advantage fit this script best. It is the most leveraged name in this scenario.

If you bet on "the elephant will eventually turn, the undervalued counterattacks" — then Samsung is the asymmetric bet: the market's expectations on its HBM are already low; once yields break through, the room for recovery is large. This is a bet that requires contrarian thinking and patience.

If you are a US equity investor and want something "directly investable with policy protection" — then Micron is the only realistic vehicle, and its "American identity" is a real plus in this geopolitical era.

And if you bet on "this cycle will eventually mean-revert, the cycle is not dead" — then you should be cautious about all three, because their high valuations already bake in optimistic assumptions of perpetual AI capex. Whoever has risen the most in this phase will fall the hardest when the tide turns.

See the pattern? "Who comes out on top" depends on which future you are betting on. This goes back to what I've said repeatedly in asset allocation: don't predict the future; instead, think clearly about what your bet means under different futures. The three giants are not a multiple-choice question of "pick one"; they are a map of "different scenarios, different odds."

066. Final Thoughts

The three memory titans are an excellent sample for observing "the same supercycle, three different ways to survive and thrive."

SK Hynix tells you about the power of focus and first-mover advantage — a #2 player, by single-mindedly betting on one track, overtook and became the leader. Samsung tells you about the paradox of scale — the giant with the deepest pockets may fall behind precisely because it turns too slowly. Micron tells you that identity can also be an asset — in an era where geopolitics reshapes supply chains, "where you make it" matters as much as "how well you make it."

But all three share one master switch, and I must emphasize this again: no matter how different their playbooks are, their prosperity rests on the same foundation — the sustainability of AI capital expenditure. As long as this well keeps flowing, all three will feast, just in different measures; once the well dries up, all three will suffer together, just at different times and depths. They are three different ships, but they sail on the same sea.

So, if I had to leave only one sentence —

Don't ask which giant will win. First ask which future you are betting on — HBM dominance, the elephant's counterattack, policy dividends, or cyclical reversion. Once you are clear on that, you'll know which of the three is your bet.

The memory series ends here. This track — the one I have dissected the most — from the overview to HBM to the three giants, has one core purpose: understanding why power shifts in an industry, and whether (and when) it will shift back. Next track, let's move further up the computing stack — the AI compute stack itself: from GPU to network to power. Money is flowing along this stack; let's see where it goes.

——

Risk Disclaimer: This article is an industry study. The companies mentioned are for analysis purposes only and do not constitute any investment advice. Market risk exists; invest prudently.