Serenity spent Tuesday pushing back against what he characterized as one of the sloppiest thematic selloffs of the year, as a wide swath of AI-linked names dropped anywhere from 4% to double digits.

The selloff and his read. He tied the drop to three catalysts he considers hollow: a Meta (META) compute story he says contained nothing new, and two separate reports of delays to co-packaged optics — CPO, the technique of placing the optical engine next to a switch or GPU die to cut power and boost bandwidth — both of which he noted were refuted by NVIDIA (NVDA). NVDA sits at the center of his framework as the demand signal he uses to front-run upstream suppliers rather than a name he owns. Serenity argued the damage was driven by cascading margin liquidations rather than fundamentals, pointing to a cluster falling together — neocloud NBIS, chipmakers MRVL, INTC and AMD, memory names SNDK and MU, and his photonics longs SIVE (indium-phosphide lasers, his largest position) and LITE (Lumentum, his large-cap optics incumbent). His view: indiscriminate drops on improving underlying names usually create opportunities, and institutions will likely end up buying the same names retail is capitulating on.

Memory and the Samsung datapoint. He flagged that Samsung — a proxy for his DRAM/NAND supercycle thesis, the idea that AI demand is forcing sustained memory price hikes — posted results he framed as the world's most profitable, hiking DRAM roughly 20% and growing operating profit sharply, yet still sold off. He treats this as confirmation the market is misreading the memory cycle, which underpins his MU and SNDK longs.

Late damage control. Serenity noted Meta later stressed it remains hungry for compute and cited fresh deals with names including Oracle (ORCL) and a cloud partner, calling it too-late cleanup after the earlier framing had already triggered liquidations. He treats Oracle mainly as a hyperscaler-demand datapoint rather than an investment.

LiDAR and lasers. A potentially bullish development for his supply-chain names: Chinese LiDAR maker Hesai faces US scrutiny over its NVDA partnership, with warnings that its sensors could be remotely disabled. Serenity argued that regulating out a Chinese competitor helps Western LiDAR players OUST and his speculative 4D FMCW LiDAR long AEVA, and — by second-order effect — the upstream laser suppliers those systems use, namely LITE and SIVE. AEVA already sources continuous-wave lasers from Sivers, reinforcing the chokepoint link.

Odds and ends. He mocked a Raymond James initiation on SpaceX (SPCX) at an ~$10T implied valuation as a retail-facing note dressed up as institutional research. On robotics, he cited Chinese projections of 100,000-plus humanoids this year versus far lower Western estimates, calling for more US support and naming Tesla (TSLA) and Agility-linked CCXI as catch-up candidates — framing it as a possible "Sputnik moment" for robotics that would pull demand through upstream supply chains and rare earths.

He also took swipes at chart-based traders claiming to have called moves like AXTI's past run, arguing valuations in names like AXTI, SIVE and test-equipment play AEHR are set by InP demand, hyperscaler orders and volume ramps — not technical patterns.

*This is derived commentary, not investment advice.*