Serenity, the small-cap-focused stock picker whose reconstructed portfolio this site tracks, spent Friday returning to one of his highest-conviction stories: the idea that AI is driving a structural, multi-year shortage in computer memory rather than a passing price blip.

The trigger was a research call from a Meritz Securities analyst arguing that the market is badly misreading the semiconductor memory setup and that this is the wrong moment to sell Korean memory makers. Serenity highlighted the specifics: on that view, DRAM suppliers will only be able to meet roughly 75%–80% of demand in the second half of 2026, with fulfillment slipping into the 60% range in 2027. DRAM is the workhorse memory used across servers and PCs, and its high-bandwidth variant (HBM) is stacked alongside AI GPUs — a packaging bottleneck Serenity has flagged repeatedly. The same note pointed to strikingly low forward valuations for the Korean players, with SK Hynix (SKHY) and Samsung trading at low-single-digit multiples on 2027 earnings, a discount Serenity reads as evidence the market still doesn't believe the shortage is real.

SK Hynix matters to his framework as one of the three dominant memory manufacturers worldwide alongside Samsung and Micron. He typically treats the Korean names as leveraged proxies on the same supply crunch, while favoring the US supplier on tariff and policy grounds.

That US supplier, Micron (MU), remains the centerpiece of his "memory supercycle" thesis and one of his largest disclosed positions. Micron is the only major American maker of DRAM and NAND flash, and Serenity has long argued its "made in America" status is a structural advantage over its Korean rivals if tariffs or export frictions bite. Today he pointed to Micron reportedly signing 16 long-term agreements featuring favorable take-or-pay volume commitments — contracts where customers lock in and pay for supply well in advance — as concrete confirmation that demand is being pre-booked years out rather than merely hoped for. That, in his telling, is the difference between a cyclical bounce and a durable re-rating; he has modeled Micron gross margins in the mid-70s and framed the stock as an underappreciated re-rating candidate.

He also folded in a comment attributed to the Intel (INTC) chief executive suggesting there will be no meaningful relief in memory supply or pricing before at least 2028 — a timeframe that lines up neatly with his own view. Intel is a separate, policy-driven holding for Serenity, framed as a bet that Washington won't let its domestic foundry champion fail; here he simply used the CEO's remark as an outside voice corroborating how long the squeeze could run.

Nothing in today's activity signaled a new position or a change in weighting. Instead, it was a conviction-reinforcing day: Serenity gathering third-party data points — a Korean sell-side call, Micron's contract book, and an Intel timeline — to argue that the memory shortage he has bet on is structural, not transitory.