Monthly Newsletter #5: The Quiet Repricing

Welcome to this month's edition of what I've been thinking about in the technology space. Rather than recap market movements, this issue takes two sides of the same question: where pricing power is quietly emerging in commodity tech, and how to read the margins it produces.

When Commodities Stop Being Commodities

Commodity typically means interchangeable: customers can swap one product for another without trouble. But AI servers are de-commoditizing certain parts of the supply chain, driving up both the specs and the volumes. Only a handful of suppliers can keep up.

Take capacitors: as chip power density increases, they must deliver higher capacitance at tighter tolerances to fill the gap when a chip spikes from idle to peak. A single AI server uses 10-15x more capacitors than a traditional server, and each unit commands prices up to 10x higher than consumer-grade parts. The few suppliers who qualify are raising prices accordingly.

Or take circuit boards. These must carry ultra-fast signals while managing heat from chips drawing over a kilowatt each, pushing every material to its limit: insulating resin, copper foil, glass yarn. Each is dominated by a quiet oligopoly. And mastering each material is only half the battle; laminating them into one board is the other half, matching how they expand and contract with heat. An industry that once counted 2,800 fabricators has thinned to a handful that qualify for AI-grade work.

Physics is pushing legacy supply chains past their limits, rewriting the rules of what was once commodity tech.

Margin Is Not the Message

Every layer of the AI infrastructure stack is running short on supply. Yet some suppliers swing from deep negative to 80% margins within two years, while others accrete just a few points annually. The question every investor asks is which layer will see the biggest margin expansion. We don't think gross margin alone can answer it.

The game depends on the market structure. In commodity markets, many suppliers face many anonymous buyers. Marginal scarcity dictates pricing, so cyclical players optimize accordingly: maximize cash during booms to survive the busts. In non-commodity markets, few suppliers face few customers, often co-designing products together. The incentive shifts from surviving cycles to growing long-term TAM. Margins stay steady rather than swinging wildly.

Either way, capital runs its own cycle. High profits attract capital, which expands capacity, invites competitors, and drives returns back down. Poor profits repel capital, forcing consolidation until discipline returns. Consider memory: the previous down cycle left players underinvested. Now astronomical returns are driving massive capacity expansion, while inviting aggressive competition from China. All that capex becomes depreciation, a fixed cost that arrives regardless of where demand goes.

The same high margin can precede a capacity flood, a decade of scarcity pricing, or obsolescence. The right question is not how high the margin could go, but what game is being played, what behavior it will provoke, and where we are in the cycle.

What do you see as the most underweighted topic in tech for 2026? As always, I'd welcome your thoughts on any of this.

Until next month,

Jess Xu