Serenity, the stock-picker whose reconstructed portfolio this site tracks, kept his commentary brief but pointed on Sunday, weighing in on a pair of headlines that speak to the scale of money now sloshing through the AI infrastructure buildout.

## Nvidia's circular financing

The bigger of the two items concerned NVIDIA (NVDA), the dominant designer of the graphics processors that power AI training and the company Serenity treats as the central reference point of his entire approach. He does not own the stock; instead he uses NVDA's roadmap and spending as a demand signal, then hunts the smaller, upstream "chokepoint" suppliers — the lasers, substrates and packaging tools with few substitutes — that its buildout makes scarce.

Serenity relayed reports that NVDA is in talks to backstop roughly $250 billion of financing tied to OpenAI's data-center expansion. The arrangement, as he described it, would see SoftBank's energy arm develop a 10-gigawatt campus in Ohio, OpenAI sign a long-term lease, and NVDA guarantee the quarter-trillion in financing — all alongside a separate discussion for OpenAI to buy up to $350 billion in Nvidia chips.

He was candid about the optics: a chip vendor guaranteeing the financing that funds purchases of its own chips is the kind of circular arrangement whose long-term health he questions. But he set that skepticism against the sheer size of the AI infrastructure spend, and his takeaway was consistent with his framework — whatever one thinks of the mechanics, the beneficiaries downstream and upstream of this capex should be very well fed. That is the logic behind his positioning in optical interconnect suppliers, memory names and packaging chokepoints, all of which he views as leveraged to exactly this scale of committed spending.

## A frothy China listing

The second item was more of an aside. Serenity noted the trading debut of CMXT, a newly public Chinese name whose shares he flagged as having surged nearly 470% on its first day, lifting its valuation to roughly $487 billion from about $85.5 billion. He offered no thesis on the company and asked, half in jest, whether any of his China-based followers had gotten into the IPO.

The remark reads less as a recommendation than as a wry marker of just how much speculative appetite is chasing anything adjacent to the AI theme — a mood that cuts both ways for a deep-value investor who prefers obscure, under-owned suppliers to crowded, richly valued headline names.

## The takeaway

Nothing in Sunday's posts signaled a change to Serenity's holdings. The through-line was validation of his core premise: that the money committed to the AI buildout is now large enough that the upstream links he favors — the light-source and substrate suppliers feeding co-packaged optics, the memory and advanced-packaging bottlenecks — sit in the path of enormous, telegraphed demand. He continues to argue the better risk-reward lies in those quiet chokepoints rather than in the mega-cap at the center of the story.

*This is derived commentary, not investment advice, and reflects a paraphrase of Serenity's public posts.*