Nvidia’s enormous supply-and-capacity commitments provide the central clue in an argument that the AI infrastructure bottleneck may be hiding somewhere less fashionable than power generation. The presentation points to Nvidia’s commitment figure rising from $119 billion to $279 billion and emphasizes the company’s stated focus on memory, manufacturing facilities, and infrastructure. From there, it asks a useful investing question: rather than simply finding companies enjoying booming demand, which suppliers possess enough scarcity to keep a greater share of the economics for themselves?
Gross-margin expansion becomes the creator’s primary test for answering that question. The logic is intuitive: revenue can surge because an entire industry is busy, whereas expanding margins can indicate that a supplier has unusual pricing power. Comparing the most recent four quarters with the preceding four gives the discussion a consistent framework, and applying it across 24 companies is more informative than selecting one or two convenient examples. Still, gross-margin expansion is only a proxy for scarcity. Product mix, input costs, operating efficiency, accounting changes, cyclical recovery, and other factors can affect margins, so the test cannot by itself establish exactly why profitability changed.
The comparison between power-and-cooling suppliers and memory companies is particularly effective. Vertiv is presented as a legitimate beneficiary, with revenue growth, rising deferred revenue, inventory expansion, and a 4.3-point gross-margin improvement supporting the case that demand is substantial. By contrast, nVent’s reported 2.2-point margin decline despite 46% revenue growth is used to argue that rapidly expanding sales do not necessarily translate into exceptional supplier leverage. The creator occasionally pushes the interpretation too far—customers prepaying billions does not automatically prove that a supplier is irreplaceable, for example—but the broader distinction between participating in a boom and controlling its scarce input is worthwhile.
Memory produces the dramatic results. SanDisk is described as gaining 41.4 gross-margin points while revenue nearly tripled, while Micron is said to have moved from a 37.7% quarterly gross margin to 84.6%, alongside long-term supply agreements and customers reportedly receiving only part of the memory they want. Those figures, if interpreted as presented, make a strong case that memory suppliers currently possess substantial pricing power. The discussion is also strengthened by finding similar behavior in different types of memory rather than treating Nvidia’s direct purchasing relationships as the sole evidence. However, the extraordinary profitability itself creates a problem: unusually high margins can signal tremendous scarcity today without proving that those economics are sustainable.
That cyclicality receives serious attention rather than being buried is one of the analysis’s better features. Micron’s historical swing from a previous margin peak to eventually losing money provides a clear warning about extrapolating peak conditions, while the creator explicitly notes that Micron, Vertiv, and nVent rank near the top of their respective valuation histories under his methodology. His claim that Micron trades around six times the coming year’s earnings sounds strikingly inexpensive, but a forward multiple based on exceptionally elevated expected earnings can be deceptive if the cycle reverses. The stated fair-value discount and margin-of-safety conclusions therefore need considerably more underlying modeling than viewers are shown before they can be treated as persuasive valuation evidence.
China supplies a credible challenge to the bullish memory thesis and prevents the presentation from becoming a simple stock pitch. The discussion considers a heavily state-funded Chinese competitor adding substantial wafer capacity, potential Apple qualification, political opposition in Washington, and the distinction between wafer starts and actual market share. More importantly, the creator identifies additional commodity-memory capacity as exactly the kind of development that could weaken current scarcity. Connecting that risk with reported swings in Micron’s share price makes the uncertainty tangible, although short-term stock reactions cannot establish that the market agrees with a particular interpretation of those political developments.
The broader comparison ultimately makes the thesis more interesting than a straightforward memory-stock recommendation. Constellation Energy emerges as an exception to the argument that electricity infrastructure is difficult to own as a bottleneck because existing, licensed nuclear generation cannot simply be manufactured quickly, and the creator appropriately stops short of recommending it without completing his valuation work. Less convincing is the sponsored Copper One Resources segment: although risks and the company’s early-stage status are acknowledged, the lengthy promotion of a tiny exploration company interrupts an otherwise systematic examination of established AI-infrastructure businesses. Overall, the presentation offers a clever framework for investigating scarcity, but viewers should treat gross-margin changes as a starting signal for deeper research rather than a standalone method for identifying the winners of the AI buildout.
Pros
- Builds the analysis around a clear, repeatable comparison of gross-margin changes rather than relying only on revenue growth or AI enthusiasm.
- Uses Nvidia’s stated supply-and-capacity commitments as a concrete starting point for investigating where infrastructure spending is flowing.
- Compares memory, cooling, electrical equipment, power generation, and other infrastructure categories instead of examining only the companies that support the preferred thesis.
- Takes memory cyclicality and expanding Chinese capacity seriously as potential threats to exceptionally high current profitability.
- Distinguishes between owning an industry bottleneck and merely benefiting from strong demand, a useful concept for investors evaluating AI infrastructure companies.
- Acknowledges important limitations in the creator’s own valuation system, including SanDisk’s short public history and the absence of completed valuation work on Constellation Energy.
Cons
- Treats gross-margin expansion as stronger evidence of scarcity and pricing power than the metric alone can establish.
- The Micron and SanDisk valuation conclusions are not supported with enough visible financial modeling to evaluate the stated fair values and margins of safety.
- Forward earnings multiples can look unusually attractive near a cyclical profit peak, making Micron’s roughly six-times figure less decisive than the presentation sometimes suggests.
- Several categorical statements—particularly about customer prepayments, supplier replaceability, and what factories can readily reproduce—simplify more complicated competitive dynamics.
- The sponsored Copper One Resources section is lengthy and sits awkwardly inside an analysis supposedly designed to identify bottlenecks through a consistent financial test.
The scarcity framework is a thoughtful way to separate AI businesses that are simply enjoying huge demand from suppliers that may possess genuine pricing power, and the memory results make a compelling case for further investigation. The argument is strongest when it compares companies using consistent financial measures and weakest when those measures are treated as sufficient proof of durable competitive advantage or attractive valuation. With its unusually serious treatment of cyclicality and competing capacity, this is useful investment analysis, but the stock conclusions still require substantially more financial work before following them with capital.
