Serenity, the stock-picker known for front-running the AI buildout by buying the obscure upstream suppliers — lasers, substrates, memory — that the big names quietly depend on, spent Thursday parsing fresh earnings and news for confirmation of his core theses.

Tesla as a supply-chain telegraph. Serenity combed through Tesla's (TSLA) earnings transcripts, flagging Elon Musk's repeated, high-conviction framing of the Optimus humanoid robot as the company's biggest product ever. He does not own Tesla — he treats it as overvalued on conventional metrics — but uses it as the anchor for his robotics supply-chain framework, preferring to play the Optimus ramp through under-the-radar component suppliers rather than the automaker itself. He seized on remarks from a Tesla supply-chain executive about building new supplier bases from scratch for memory, metal-injection-molded parts and flexible printed circuits, reading it as a map of where fresh chokepoint opportunities might emerge. He also noted Musk thanking Micron for a significant memory allocation on reasonable terms despite soaring prices.

Long-term agreements everywhere — his favorite tell. The bigger theme Serenity drew out was the spread of multi-year, take-or-pay long-term agreements (LTAs) across the supply chain. He pointed to a report that Intel (INTC) and AMD are signing CPU LTAs with Chinese AI data-center customers, with some CPU prices up more than 40% in China this year. Intel is his policy-driven bet that Washington won't let its domestic foundry champion fail; AMD he doesn't own but treats as a demand engine pulling through his real upstream picks. He tied the CPU squeeze — where AI inference has made server CPUs a bottleneck — to parallel LTAs elsewhere: in memory, where Micron (MU) and Sandisk (SNDK) plus the Korean giants are locking in DRAM and NAND deals, and in photonics, where Lumentum (LITE) and Coherent (COHR) are signing agreements for EML lasers (the electro-absorption-modulated light sources behind 800G/1.6T links).

For grounding: Micron and Sandisk are the twin pillars of his "memory supercycle" thesis — a structural, AI-driven shortage of DRAM, HBM and NAND that he argues supports repeated price hikes and multi-year demand out to roughly 2028. Lumentum and Coherent are his quality large-cap photonics longs, benefiting from the optical-interconnect buildout. He highlighted a notable new wrinkle: reports that AMD and hyperscalers are now pursuing LTAs for CW (continuous-wave) lasers — the light source for silicon photonics and co-packaged optics (CPO), where the optical engine sits next to the switch or GPU die. That matters to his framework because CW-laser demand feeds directly into the smaller upstream names he favors as CPO suppliers.

The takeaway. Serenity framed the day's news as a point in favor of what he calls bottleneck investing. His argument: it is hard to dismiss the AI trade as a bubble poised to pop when demand is being locked up years in advance through take-or-pay contracts spanning memory, CPUs, substrates and photonics. He also singled out TSMC (TSM) as the indispensable foundry underpinning the whole buildout — a safe anchor he holds while chasing higher-upside upstream names.

*This is derived commentary, not investment advice, and paraphrases Serenity's posts rather than his exact words.*