The AI-driven memory supercycle is no longer a forecast — it’s a financial reality reshaping the global semiconductor landscape. SK Hynix and Micron have both posted earnings that would have seemed like science fiction just two years ago, but beneath the headline numbers lies a more complex story about competitive positioning, HBM dominance, and what comes next. We asked our three analysts to weigh in.
Let me be precise about what we’re actually looking at here, because the euphoria in the market is getting dangerously ahead of the fundamentals. Micron’s fiscal Q3 2026 results were objectively extraordinary — revenue and operating profit came in roughly 16% and 21% above consensus, respectively, with EPS of approximately $25 representing year-over-year growth of over 1,200%. That is not a typo. SK Hynix, meanwhile, is projecting operating profits that some sell-side analysts are penciling in at 27 to 30 trillion Korean won for Q1 2026 alone. These are historic numbers by any measure.
But here is what concerns me structurally. Micron itself flagged in its guidance commentary that while the next quarter looks strong — EPS guidance of around $31 — the rate of memory price appreciation is expected to slow meaningfully. We are likely near peak price velocity, not peak prices themselves, but the distinction matters enormously for how you model forward earnings. The memory industry has a well-documented history of confusing cyclical peaks with secular shifts. AI demand is real, but the question I keep returning to is whether the current supply expansion — Samsung alone is reportedly ramping HBM4 DRAM wafer capacity by 120,000 wafers per month, with some reports suggesting capacity could reach 150,000 wafers — will eventually create the same oversupply dynamics we have seen in every previous memory cycle. The HBM market is not immune to physics or economics.
Additionally, the macro backdrop for Korea specifically warrants caution. SK Hynix’s planned U.S. ADR listing, targeting roughly 45 trillion won in capital raises, signals that management itself is seeking dollar-denominated capital at scale. That tells you something about where they see currency and liquidity risk. When a Korean champion hedges its financing into USD, the macro bear in me pays close attention.
I’ll be honest — I spent years being skeptical of HBM as an investment thesis. It looked like a niche product solving a niche problem. What changed my view was not the AI hype cycle, but the actual financial architecture that SK Hynix has built over 14 years of what amounts to disciplined, contrarian capital allocation. When SK Group acquired Hynix in 2012, the deal was widely criticized. The company was debt-laden, operationally struggling, and competing in a commodity market. The decision to double down on high-bandwidth memory when it was genuinely a niche product — years before anyone was seriously talking about AI training at scale — is the kind of capital allocation decision that separates great businesses from average ones.
The numbers now bear this out in a way that is almost uncomfortable to look at if you missed it. On a comparable basis, SK Hynix’s operating margins are converging toward — and in some periods exceeding — those of Micron, which itself is now generating gross margins that were historically associated with fabless semiconductor companies, not DRAM manufacturers. What I find particularly interesting about Micron’s recent results is the emergence of what they are calling Strategic Customer Agreements — long-term contracts that, when fully executed, will lock in margins on roughly half or more of Micron’s revenue across multiple end markets. This is a structural change in how memory is being priced and contracted, and it represents genuine intrinsic value creation, not just cyclical margin expansion.
Samsung’s HBM position is the wild card in my valuation work. Reuters reporting suggests Samsung may be the first of the three memory majors to begin mass production of HBM4 for Nvidia — though the announcement was not official as of the reporting date, and Korean media has a documented tendency to sharpen the tone of what are still tentative supply chain signals. If Samsung genuinely closes the HBM quality gap with Hynix at the HBM4 generation, the competitive moat I am ascribing to Hynix narrows. I am watching this very carefully before adjusting my long-term models.
Look, I don’t need to read a 40-page sell-side report to tell you what’s happening here. Micron just delivered one of the greatest quarterly earnings prints in semiconductor history, and the market moved. SK Hynix is listing ADRs on Nasdaq — this is a $45 trillion won capital event that is going to force every major U.S. institutional investor to build or expand a position. That’s not a thesis, that’s a calendar event with money attached to it. When you combine a landmark earnings beat from Micron with a forced-buying dynamic from the Hynix ADR launch scheduled for July 10th, and you throw in analyst upgrades cascading through Samsung Electronics’ Q2 estimates, you have a momentum setup that is hard to fight.
The detail I keep coming back to from the Micron print is the NAND angle. NAND revenue nearly doubled quarter-over-quarter — a near 100% QoQ increase — and that part of the story is getting almost no coverage because everyone is fixated on HBM. But that NAND inflection tells me the memory recovery is broader and more durable than a single AI chip story. This isn’t just Nvidia buying HBM. This is enterprise storage, traditional servers, and AI inference all pulling inventory simultaneously. The HPE and Dell server results confirmed this — networking revenue up triple digits year-over-year, and the stated constraint is supply, not demand. The customer is accepting price increases without flinching.
On the HBM competition specifically — Hynix won the last generation, Samsung is fighting back in HBM4, and Micron is quietly investing $200 billion in domestic U.S. capacity while enjoying the geopolitical tailwind of being the American champion. That’s three different stories running in parallel. For a trader, that means this theme has multiple catalyst legs left. For anyone sitting on the sidelines waiting for a clean pullback to enter memory names — you might be waiting a long time.
The SK Hynix and Micron earnings debate ultimately reveals a memory market in genuine structural transformation — not merely another cyclical bounce. The Macro Bear is right to flag that price appreciation rates will slow and that capacity additions are accelerating at a pace that deserves scrutiny. The Value Hunter is right that the long-term contrarian bets — particularly Hynix’s 14-year HBM commitment and Micron’s shift toward contracted, sticky revenue — represent real intrinsic value creation. And the Street Pragmatist is right that right now, the momentum, the capital flows, and the fundamental backdrop are all pointing in the same direction. The HBM competition between Hynix, Micron, and a resurging Samsung will define the next chapter of global AI infrastructure — and Korea’s position at the center of that competition is no longer a bet, it’s a fact.