Why Samsung and SK Hynix Are Winning the Memory Supercycle

Micron Just Proved What Samsung and SK Hynix Knew All Along

Micron reported fourth-quarter revenue of $54.23 billion for its fiscal 2026 — roughly four times its year-ago figure, and well above Wall Street’s expectations. Adjusted earnings per share hit $33.42, also beating estimates. The numbers are a vindication of the memory supercycle thesis that has been building since 2024, and they landed in South Korea on a day when the KOSPI closed at 6,971.35, up nearly 2 percent.

But the headline number isn’t Micron. It’s what Samsung Electronics and SK Hynix did while Micron was announcing its results.

Samsung rose 2.23 percent. SK Hynix gained 2.82 percent. More importantly, domestic institutions bought 334.3 billion won worth of shares that day — flipping from sellers to buyers in a single session after months of caution. Individuals and foreigners, by contrast, offloaded 1.43 trillion won and 543.7 billion won respectively. That’s a structural shift: local capital is committing to a sector that foreign investors have been rotating out of.

What this divergence reveals is a growing belief gap. Foreign holders, many of whom piled into Korean semis during the 2023–2024 recovery, appear to be taking profits ahead of potential cyclicality. Korean institutions, watching the same data but reading it differently, see something more durable. The question now is whether that conviction holds when the next earnings season tests it.

The Real Winner Is DRAM Pricing Power

Western coverage of Micron’s earnings has been dominated by the AI narrative — how high-bandwidth memory is feeding GPU clusters, how data centers are absorbing supply. What gets glossed over is the pricing mechanism underneath all of it.

DRAM prices have been climbing since mid-2024, and Micron’s results confirm they’re still climbing. Samsung and SK Hynix are responding differently from Micron in one key way: they are not just reacting to demand, they are managing supply more aggressively. Samsung has been deliberately capping production growth in its memory divisions to protect margins. SK Hynix has been prioritizing HBM — high-bandwidth memory — over standard DRAM, which carries higher margins and is in structurally tighter supply.

The implication for pricing through 2028 is significant. DRAM is not returning to the price floors of 2022–2023. Even if AI demand normalizes, the cost structure of building new fab capacity and the lead time required to bring it online means supply cannot surge quickly enough to wipe out price gains. That is the definition of a supercycle — prolonged elevated pricing, not a temporary spike.

Beyond the pricing story, there’s a second-order effect playing out in capital allocation. Samsung’s memory division is now generating enough free cash flow to fund aggressive R&D in next-generation nodes while maintaining shareholder returns. SK Hynix is doing the same, but with a sharper focus on HBM3E and the upcoming HBM4 standard. Both companies are effectively using supercycle cash to widen the moat between themselves and any potential entrants — including Chinese memory producers attempting to catch up with state-backed investment.

The Supply-Side Lock-In

One detail deserves more attention than it’s received: the lead time for new memory capacity remains stubbornly long. Building a new fab or even retooling an existing line for advanced DRAM processes takes 18 to 24 months from decision to volume production. Samsung and SK Hynix understood this before Micron’s earnings call. They have been making capital decisions based on demand visibility extending well into 2027, not just the current quarter.

This creates a feedback loop. High prices justify capital spending, but that spending doesn’t translate into supply for two years. By the time new capacity comes online, the demand picture may have shifted again — or the companies may choose to slow down rather than flood the market. Micron’s management has publicly acknowledged this dynamic, noting that the company is being deliberate about ramping production to avoid repeating the glut cycles that punctuated the 2010s.

The net result is a pricing floor that didn’t exist in previous memory cycles. In the past, a price increase would trigger rapid supply response and a quick reversion. Today, the economics of fab construction and the strategic priority of HBM over commodity DRAM are creating a floor — and a ceiling — that are both higher than before.

Who Wins. Who Loses.

The winners are clear. Samsung and SK Hynix are capturing margin expansion that their U.S. peers are sharing but not leading. Samsung’s foundry business also benefits indirectly: advanced memory packaging requires cutting-edge fabrication, and Samsung is investing heavily there. SK Hynix is the closest supplier to NVIDIA’s HBM pipeline, which gives it pricing leverage that extends beyond any single earnings cycle.

But the winners list has an entry that rarely appears in these discussions: Korean equipment makers. As Samsung and SK Hynix continue expanding and upgrading their memory fabs, companies like Samsung Electro-Mechanics, LS Mtron, and HD Semiconductor are seeing order books fill with capex that won’t be delivered for quarters. This is the quiet multiplier of the supercycle — a downstream boom that follows the upstream headline numbers.

The losers are the players who assumed memory pricing would revert. Investors who bought Korean semiconductors on the hope of a price correction will be waiting a long time. And companies at the other end of the supply chain — cloud providers, server manufacturers, even smartphone makers — are absorbing higher memory costs that will ripple through hardware margins through at least 2027. Apple’s decision to stick with 8GB base RAM in newer iPhone models, for example, isn’t just product strategy. It’s cost management in a market where memory is no longer a cheap input.

There’s also a geopolitical angle. China’s state-backed memory ambitions, led by ChangXin and YMTC, remain constrained not just by technology but by U.S. export controls on equipment. Samsung and SK Hynix are benefiting from a regulatory environment that slows their largest potential competitor from catching up on pricing pressure.

Why This Matters Outside Korea

The KOSPI closing near 6,970 is a local event. The memory supercycle is not. Samsung and SK Hynix together account for roughly two-thirds of global DRAM capacity and a dominant share of HBM. When they raise prices, the entire semiconductor supply chain feels it. When they manage output, Apple’s iPhone memory options, Microsoft’s cloud infrastructure, and Tesla’s Dojo supercomputing all adjust their cost models accordingly.

The market’s rotation matters too. Institutions buying 334 billion won in a single session suggests domestic capital is re-rating Korean semiconductors from cyclical plays to structural growth positions. That’s a signal worth watching. If foreign investors continue selling while institutions accumulate, it creates a divergence that could either validate the thesis — if earnings hold — or expose it, if the cycle turns sooner than expected.

For global investors, the takeaway is straightforward: the memory complex is no longer a commodity bet. It’s a concentrated oligopoly with pricing power, strategic supply management, and a moat built on both technology and scale. Samsung and SK Hynix aren’t just riding the wave. They’re shaping it.

What Comes Next

The next catalyst is likely Samsung’s own earnings, which will show whether its margin expansion is keeping pace with Micron’s. SK Hynix’s production guidance for HBM will indicate how much of the demand tailwind is locked in. And Micron’s commentary on inventory levels and pricing negotiations will set the tone for the rest of the industry.

But the real test will come in the second half of 2027. That’s when the capital decisions made today start producing physical capacity. If demand holds — and there’s no guarantee it will — the companies that managed supply best will walk away with margins that look ordinary today but extraordinary in retrospect. If demand softens, the same discipline will determine who survives with their pricing power intact and who is forced back into a race to the bottom.

The supercycle is real. The question now is how long it lasts — and whether Samsung and SK Hynix can sustain their pricing power when the next wave of competition arrives.