It was a reflective, defensive Saturday for Serenity, the stock-picker whose approach centers on buying the small, upstream "chokepoint" suppliers — the lasers, substrates and packaging tools — that feed the broader AI buildout. Rather than announcing trades, he spent the day answering critics and disclosing the pain in his own book.

Cultural divide, and a candid drawdown number. Serenity contrasted his growth-oriented, technology-shift approach with what he described as a capital-preservation mindset common among some sidelined European accounts — investors he says treat positions in optics or memory as gambling and cheered the recent AI selloff. He was blunt about the cost of his own concentration and leverage: running roughly 1.4x margin against a book heavy in memory and photonics, a ~35% average drawdown in the underlying names translates to something near a 49% portfolio drawdown. His posture, though, is unchanged — he argues demand across compute, energy, memory and networking is structural and expects the AI names to recover. He also took a shot at paid technical-analysis newsletters that claimed to predict the semis crash, arguing that anyone who truly believed it would have made a fortune on puts rather than selling subscriptions.

The names he defended. He grouped several of his core longs as recent casualties of the selloff after big prior runs. AAOI (Applied Optoelectronics) is his Made-in-America optical transceiver bet — a vertically integrated Texas laser-and-assembly shop he frames as sold out to hyperscalers on 800G/1.6T demand and a candidate for a large re-rating if its capacity ramp delivers. MRVL (Marvell) is his custom-ASIC and co-packaged-optics play (CPO being the move to put the optical engine right next to the switch or GPU die), notable to him mainly as the supply-chain hub linking down to his upstream laser picks. NBIS (Nebius) is his favored "Neocloud" — an AI cloud operator he values as a sum-of-parts story and considers the best-financed in its group. INTC (Intel) rounds out the group as his policy-driven bet that Washington won't let its domestic foundry champion fail. All four, he noted, had rallied hundreds of percent before the pullback.

The one actionable observation: an SK Hynix ADR arb. The most concrete item of the day concerned SKHY, the U.S.-listed ADRs of Korean memory maker SK Hynix — a company central to his memory-supercycle thesis as one of the world's dominant DRAM and HBM (high-bandwidth memory, the stacked DRAM sold beside GPUs) suppliers. Serenity flagged that the ADRs are trading at a premium of more than 25% over the Korean local shares, and that the two become convertible on July 29. He reasons that convertibility should open the door for arbitrage and compress that premium — potentially lifting the Korean listing or pressuring the U.S. shares. He noted only about 2.5% of the float currently sits in the ADR, leaving roughly 22.5% that could still convert, which he sees as the mechanism that closes the gap.

The takeaway from the day: no new positions, but a stock-picker publicly wearing a steep drawdown, reaffirming a structural AI-demand thesis across his memory and photonics book, and pointing to a specific, calendar-driven pricing dislocation in the SK Hynix ADR ahead of the July 29 convertibility date.

*This is derived commentary for information only — not investment advice, and not Serenity's own words.*