Bloomberg reported on Tuesday that Samsung Electronics is in the early stages of exploring an offering of American depositary receipts, having held preliminary discussions with banks without reaching a decision. Sources characterised the process as a review rather than a plan — no bank mandated, no commitment made, and a real possibility that nothing comes of it.

Samsung denied it. A company spokesperson said flatly that Samsung Electronics is not reviewing the possibility of issuing American depositary receipts.

The consensus bull case for memory is that the market is years away from saturation. On the demand side, that is almost certainly correct. It is also the wrong frame, and investors who anchor on it will be looking in the wrong direction when the cycle turns.

Memory downturns have never been caused by demand saturation. They are caused by supply growth outrunning demand growth. Those are different failure modes, and the second one can fire while demand is still compounding at twenty percent.

Commerce Secretary Howard Lutnick used a concrete-pouring ceremony at Micron’s fab site in Clay, New York, to tell Samsung Electronics and SK hynix that they will build memory fabs in the United States. Not by regulation. By envy. Micron is leading, he said, so the Koreans will feel jealous and ultimately have no choice but to follow.

The remarks landed one day before SK hynix listed American Depositary Receipts on the Nasdaq, and ten days after Samsung and SK hynix announced a combined ₩800 trillion investment program in Korea’s Honam region. The sequencing is the story. Washington watched two companies that control roughly 60% of the global DRAM market commit a generational capex cycle to their home country, and responded by asking them to commit it somewhere else.

Samsung Electronics just posted the best quarter any technology company has ever reported — operating profit up 19-fold year-over-year, comfortably beating consensus — and KOSPI fell anyway, dropping as much as 3-4% in early trading. The disconnect looks strange on the surface, but it’s a familiar pattern once you look under the hood.

The beat was already priced in. Samsung shares had run up sharply into the print, and a headline number matching (rather than dramatically exceeding) already-elevated expectations gave traders a clean exit point. When a stock has rallied hard on anticipation, even a genuinely excellent result can trigger profit-taking rather than a further re-rating.

SK Hynix formally launched its Nasdaq ADR offering this week, targeting roughly $28 billion in proceeds — a deal that would rank among the largest share sales in history, trailing only SpaceX’s IPO from last month. Pricing is expected Thursday, with trading set to begin Friday under the ticker SKHY.

The scale. SK Hynix is selling 17.79 million new shares via ADRs, with 10 ADRs representing one common share. The target was revised down from an earlier filing that sought closer to $29.6 billion, after the stock corrected in Seoul in the run-up to pricing. Even at the reduced size, it would surpass both the Saudi Aramco and Alibaba IPOs.

The chip industry just told the White House to keep its hands off memory pricing.

The Letter

SEMI, the trade association representing chipmakers and equipment suppliers, sent a letter to senior Trump administration officials warning that any government attempt to influence memory prices or production capacity would deepen the historic supply squeeze rather than fix it. The AI buildout is the driver of that squeeze, and SEMI’s ask is narrow: keep letting companies sign long-term supply agreements with customers, and extend the tax breaks already aimed at growing US output. In other words, more market mechanism, not less.

Samsung and SK Hynix unveiled a combined roughly $1.3 trillion (2,000 trillion won) decade-long investment plan for new fabs, AI data centers, and chip cluster development. Both stocks fell anyway — Samsung down over 5%, SK Hynix down over 3% on the announcement day, following an even sharper 9%+ plunge earlier in the week. The knee-jerk read: investors think the spending is reckless, a repeat of the 2018-2019 memory bust, or proof the AI trade is cracking.

The Wall Street Journal headline frames the memory shortage as a problem to be solved. It isn’t. The more accurate reading of the supply picture is that the crunch is the predictable output of a fixed production base being reallocated toward AI, and there is no near-term lever — industrial or political — that changes that math before 2027. For the three companies that own the supply, that is not a crisis. It is the most durable pricing-power setup the industry has seen in a generation.

The demand-side case for the AI buildout rests on token consumption going vertical: agentic workflows firing 10 to 20 inference calls per task, enterprise API volumes measured in billions of tokens per minute, hyperscaler revenue compounding faster than capex. That argument has a physical counterpart that rarely gets stated in the same breath. Every one of those tokens is a memory access. The token-growth curve is not an abstraction floating above the supply chain — it is the buyer standing on the other side of the DRAM and HBM order book.

SanDisk’s move from a $38.50 spinoff price to roughly $1,980 — about 5,000 percent in sixteen months — is not one rally but two stories stacked on top of each other, and the market is pricing them as if they were the same thing. Separating them is the only way to understand where the stock can go.

The Thesis

The first story is real and measurable: a NAND flash supply squeeze. AI inference has turned high-capacity flash into a constrained resource. Average selling prices per gigabyte are climbing, exabytes shipped are rising, and SanDisk has converted both into record revenue and a fiscal-2026 trajectory that Bank of America models at 176 percent growth. That is a cyclical earnings boom with unusually firm footing, anchored by multi-year contracts — five signed, three of them carrying $42 billion in minimum revenue and more than $11 billion in financial guarantees — structured so margins hold even at the price floor. This is the opposite of spot-commodity NAND, and it is what the bulls point to first.

The instinct to compare SanDisk and Kioxia is correct, but the framing usually is not. These are not two competing bets. They are one bet, expressed twice — and the wiring that connects them runs through the same factory floor.

The Thesis

SanDisk and Kioxia are both pure-play NAND flash manufacturers riding the same AI-inference storage squeeze. Both have re-rated into the mega-cap tier — roughly $293 billion for SanDisk, roughly $260 billion for Kioxia — and both are wagering that flash can climb the AI memory hierarchy and shed its commodity discount. The decisive fact is that they share the physical means of production: the Yokkaichi and Kitakami fabs that stamp out their NAND are a single joint venture, recently extended through 2034. When one company describes its market, it is describing the other’s. The useful question is therefore not which company wins, but which is the cleaner expression of an identical trade — and where the two diverge enough to matter.

An index that doubles in five months does not correct gently. On Friday the KOSPI fell 5.54%, its steepest single-session drop of the year, tripping the Korea Exchange circuit breaker after KOSPI 200 futures fell 5% and program trading was suspended. The trigger was external — Broadcom’s after-hours guidance, with third-quarter AI chip sales pegged at $16 billion, read as a soft edge on the AI narrative rather than a beat. But the velocity was domestic, and it was structural. A market that rode two stocks to a 100% gain cannot fall on those two stocks without falling harder than anyone else.

Jensen Huang confirmed at GTC Taipei 2026 on June 1 that all three major memory manufacturers — Samsung Electronics, SK Hynix, and Micron Technology — have been qualified and are already in production as HBM4 suppliers for Nvidia’s Vera Rubin AI platform. The announcement ended months of supply-chain speculation and, for Micron, reversed a narrative that had pressured the stock since March.

The qualification matters beyond the supplier list. Vera Rubin is Nvidia’s next-generation AI infrastructure platform, combining Vera CPUs with Rubin GPUs and large HBM4 memory stacks per server. It is in full production and scheduled to begin shipping in Q3 2026. HBM4 doubles the interface width of its predecessor, with the JEDEC standard supporting up to 2 TB/s per stack on a 2,048-bit bus versus approximately 1 TB/s for HBM3E. Every major hyperscaler ordering Vera Rubin systems will depend on this memory supply chain holding.