To The Moon

The honest framing first: what happened on July 16 and 17 is not the same thing as what happened in early July. Last week's selloff was profit-taking, SK Hynix listing rotation, and macro noise. This week's selloff has a specific, identifiable cause, and that cause has direct implications for the memory thesis that the prior selloffs did not.

The Bank of Korea raised its benchmark rate 25 basis points to 2.75% on July 16, its first hike since January 2023. That triggered a 6.48% KOSPI collapse, a circuit breaker halt, SK Hynix down 11.53%, Samsung Electronics down 8.23%, SKHY down 9% in US trading, MU down 8% below $910, and SNDK down 8% to $1,487.

What the Bank of Korea Actually Did and Why

The BOK's decision was not a surprise to economists. South Korea's CPI hit 3.1% in May 2026, meaningfully above its 2% target. Governor Shin Hyun-song, who took over in late April, had explicitly signalled decisive action against inflation. The economy grew 1.8% in Q1 2026, its fastest quarterly pace in more than five years, driven largely by the AI memory export boom. South Korean exports surged 71% year over year in June in dollar terms, their fastest pace since 1978. The BOK revised its 2026 GDP forecast to 2.6% and its 2026 inflation forecast to 2.7%, with core inflation expected to run above 2.4%.

This is an economy so hot from memory chip demand that its own central bank had to raise rates to cool it down. Let that sit for a moment. The AI supercycle that is the bull thesis for SK Hynix and Samsung is so strong that it is generating enough economic heat in South Korea to force monetary tightening. That is a fundamentally different situation from raising rates into a weak economy. Capital Economics now expects one more hike to 3.00% by year-end, and further tightening remains on the table.

Two structural complications made this worse than the rate move alone.

The won has weakened 2.93% against the dollar year to date, touching a 17-year low of 1,561.5 on June 5. Weaker won means imported energy costs rise, which feeds directly into domestic inflation. The BOK is partly fighting currency pass-through.

South Korea's financial regulator announced new measures targeting single-stock leveraged ETFs. These products, introduced in late May, have become amplifiers of volatility. SK Hynix and Samsung leveraged ETFs accounted for forced buying and selling that exaggerated moves in both directions. Tightening those rules removes an amplification mechanism the market had quietly come to rely on for quick entries and exits. The 2x leveraged Hong Kong product plunged 18% on the same session.

The HBM Fundamentals Have Not Changed

This is where the analysis needs to be precise, because the market is conflating two genuinely distinct risks.

The rate hike is a policy shock to Korean equity valuations. It is not a signal that HBM demand is softening, that AI data centre buildout is slowing, or that the NAND shortage is ending. The BOK raised rates precisely because the semiconductor export boom is so strong that it is driving inflation. The cause of the hike is the same thing as the bull thesis for these stocks.

SK Hynix delivered fiscal Q1 2026 operating profit of 9.7 trillion won, a record, on HBM shipments that more than tripled year over year. Revenue guidance for Q2 implies sequential growth. HBM4 is in customer qualification with NVDA's Vera Rubin platform. Micron's CEO stated two weeks ago that there is no line of sight to AI memory supply catching demand through 2027. None of that changed on July 16.

The specific numbers that matter: SK Hynix still holds roughly 50% of the global HBM market. Every Vera Rubin GPU requires HBM4 memory that only SK Hynix, Samsung, and Micron can supply at scale. NVDA's Rubin supercycle is ramping through H2 2026. The structural demand picture is entirely unchanged by a 25 basis point rate move in Seoul.

What has changed is the multiple investors are willing to pay for that structural demand picture, in a currency that is under pressure, by an issuer whose equity trades on a market now in an explicit tightening cycle. That is a valuation and sentiment reset, not a fundamental reset.

Samsung vs SK Hynix: Which One Do You Back?

The Tiger SG prompt frames this as the core question, and it is genuinely interesting because the two stocks represent different risk profiles within the same trade.

SK Hynix is the purer HBM play. Its market cap is more concentrated in the AI memory theme. It has higher revenue sensitivity to HBM pricing. It has the deeper qualification relationship with NVDA. A 50% HBM market share at an inflection in the AI compute cycle is an extraordinarily powerful position. The downside is that it is more expensive on traditional metrics and more volatile, as the 11% single-day drop demonstrates.

Samsung is the more complex picture. Samsung has been slower to qualify HBM4 with major customers and has reportedly faced yield challenges that SK Hynix has not. Its semiconductor revenue is more diversified across legacy DRAM and NAND alongside HBM, which is both a buffer and a drag. Its market cap is also supported by its consumer electronics and foundry businesses, which have their own cycle dynamics entirely separate from AI memory. At its current level following the 8% drop, Samsung is arguably the lower-risk entry into the Korean memory trade precisely because it has less pure AI memory concentration, though that same characteristic means it captures less of the upside if HBM pricing stays strong.

The honest take: if you believe the HBM supercycle has at least two more quarters of runway, SK Hynix offers more leverage to that thesis. If you want memory sector exposure with more margin of safety against the rate cycle risk, Samsung's diversification provides a partial hedge.

Neither choice is obviously correct. Both are cleaner after this week's selloff than they were two weeks ago.

The Leveraged ETF Warning Is the Most Important Paragraph Nobody Is Reading

Buried in the headlines about the rate hike and the stock drops is a regulatory change that deserves more attention.

South Korea's Financial Services Commission announced it will introduce new measures targeting single-stock leveraged ETFs. These products were introduced in late May 2026 and quickly attracted enormous retail participation. The 2x leveraged SK Hynix and Samsung ETFs became a mechanism for amplifying both entries and exits, creating a feedback loop where retail buying pushed prices higher and retail margin calls on the way down pushed them lower.

The CSOP SK Hynix Daily 2x product plunged 21.32% on July 17, consistent with leveraged product mechanics on an 11% underlying move. Those products force systematic rebalancing that amplifies the underlying stock move regardless of the fundamental view. When regulators tighten the rules on these products, two things happen simultaneously: volatility in the underlying stocks falls over time as the forced rebalancing mechanism shrinks, and the previous amplification of gains that attracted retail participants in the first place goes away.

For holders of these leveraged products specifically, this week was a lesson in asymmetric risk that the promoter material probably did not emphasise clearly enough. For holders of the underlying SK Hynix and Samsung equities, the regulatory tightening of leveraged ETFs is actually mildly positive over the medium term because it removes a source of structural volatility that had no connection to fundamental value.

The US-Listed Names: SKHY, MU, SNDK

The contagion into US-listed names reflects positioning, not fundamental read-through.

SKHY fell 9%. MU fell 8% below $910. SNDK fell 8% to $1,487. The DRAM ETF dropped meaningfully. None of these names have direct exposure to the BOK rate decision. They have indirect exposure through two channels: sentiment spillover from the Korean equity rout, and the risk-off rotation that any unexpected rate hike triggers in high-multiple growth stocks globally.

MU at below $910 is worth examining specifically. The stock was at $1,190 after its blowout earnings two weeks ago. The pullback to $910 represents a 24% retracement from the post-earnings high, on fundamentals that have not changed at all since that report. DA Davidson's $2,000 target has not been revised. Cantor Fitzgerald's $1,500 target has not been revised. The Anthropic supply agreement announced on the earnings call is still in effect. The Q4 guidance of $50 billion in revenue is still the standing expectation. The multiple you are buying at $910 is meaningfully lower than the multiple at $1,190, for the same earnings power.

SNDK at $1,487 has now retraced from its recent highs by approximately 37%. This is the stock that was at $2,354 at its peak this cycle. The structural analysis from the early July article still applies: the shortage extends to 2027, Bernstein's $3,000 target is still standing, BofA and Citi are at $2,500. The risk profile at $1,487 is genuinely more attractive than the risk profile at $2,300, though the SK Hynix listing overhang and now the Korean rate cycle risk have added genuine complexity that did not exist at the start of the year.

The Trade Framework

The BOK hike changes one thing and does not change another.

What it changes: the discount rate applied to Korean semiconductor equities. Higher rates in Korea mean institutional investors apply a higher cost of capital to future earnings streams. That is a valuation reset that is real, not imaginary. If the BOK hikes again to 3.00% by year-end as Capital Economics expects, that reset deepens. Investors in SK Hynix and Samsung need to factor ongoing rate risk into their holding thesis in a way they did not need to before July 16.

What it does not change: the AI memory shortage, the HBM demand runway, the Vera Rubin supercycle, or Micron's Q4 guidance of $50 billion. These are demand-side realities that a 25 basis point rate move in Seoul does not affect. The same NVDA chips still need the same HBM4 memory. The same hyperscalers are still building the same data centres.

The framework for each situation:

If you hold SKHY or Korean ETF exposure: this is a policy shock that may have one more leg down if a second hike comes. The HBM thesis is intact but the valuation reset is real. Trimming into any bounces to reduce rate-cycle exposure while maintaining a core position is rational. Not the same as exiting the thesis.

If you are looking at MU at sub-$910: this is the cleanest entry in the memory complex right now. US domicile means no BOK rate exposure. Fundamentals just printed a $41 billion quarter. The 24% retracement from earnings highs with no change in the underlying story is the definition of a better entry. The risk is that further Korean contagion or Warsh Fed hawkishness adds another 10 to 15% downside before the fundamental gravity reasserts.

If you are looking at SNDK at $1,487: more complex than MU given the SK Hynix listing overhang that has not resolved. But a 37% retracement from cycle highs on unchanged fundamentals is a meaningful improvement in risk-reward versus buying at $2,300.

The memory bloodbath is a policy shock landing on a genuinely strong fundamental story. Those two things are not in conflict. The AI boom is so powerful it caused a rate hike in the country that supplies most of the world's memory. The rate hike then caused the stocks of the companies creating that boom to fall 8 to 12% in a single session. That is the paradox sitting at the centre of this trade right now, and it does not resolve until either the BOK stops hiking or the AI demand story weakens. Neither of those events appears imminent.

I am not a financial advisor. Trade wisely, Comrades.