The Latent Lever

Steve Williams · babnews.org

The whole market is busy sorting the AI buildout by who has pricing power. It's the right question and it's become a crowded one. The more interesting version, the one with alpha still in it, is the inverse: who holds real pricing power and hasn't used it yet, and why not. Because the "why not" is where an operator sees something a stock screen structurally cannot. A screen reads a company's current pricing as its ceiling. An operator who has sat inside these businesses knows that a lot of the time the current price is a choice, not a limit, and the reason behind the choice tells you exactly when it breaks.

Latent pricing power is not the same as cheap. A cheap stock might just be fairly priced for a weak business. Latent pricing power means a company structurally could charge more or capture more margin and is choosing not to, and the choice has a cause you can name. Sort by the cause, and a list of stocks becomes a map of triggers. Here are the five reasons the lever sits unpulled, each with a rung of the stack where it's live, the operator's read on it, and what flips it.

1. Scar tissue

The rung: the sub-tier suppliers. Vacuum valves are roughly three-quarters held by one company. Silicon wafers are a Shin-Etsu and SUMCO duopoly. The build-up film under nearly every advanced package is north of 95 percent one supplier. These are monopoly and duopoly positions, and they price like companies bracing for a downturn, modest increases, cautious capacity.

The operator's read: this restraint is trauma, not weakness. Every one of these firms has been whipsawed by prior semiconductor cycles, boom orders followed by cancellations and idle capacity they paid to build. Having been burned repeatedly, they price and invest defensively, fearing the next bust more than they covet this boom. A screen reads the conservative pricing as the company's ceiling. An operator knows it's a posture, and postures change.

What flips it: belief, not demand. Demand is already here, order books full and lead times doubled. The repricing comes when a monopolist decides the cycle is structural rather than another head-fake, and because there is no second source, the move is violent when it comes.

2. Contracts signed before the scarcity

The rung: anything sold on multi-year agreements. Long-term wafer supply deals, substrate contracts, and especially power purchase agreements were often struck when the input was abundant and cheap. The supplier holds the leverage now but is honoring a price set in a looser world.

The operator's read: the pricing power is real and already earned; it's simply time-locked. You can almost put a date on its release, the renewal. This is the most predictable form of latent power, and the least appreciated, because the income statement today shows the old price and the screen extrapolates from it.

What flips it: the contract cycle. Watch renewal calendars, not spot prices. The repricing is scheduled; it just hasn't arrived on the P&L.

3. Fear of the customer

The rung: the customer-concentrated designers and assemblers. Some suppliers with genuine switching costs sell into a handful of giant buyers who could, in principle, design them out or bring the work in-house. So they underprice to protect the relationship.

The operator's read: this is the subtlest case, because the restraint is rational and the power is self-suppressed rather than absent. The tell is whether the switching cost is real. If re-qualifying away from this supplier would cost the customer a year and a yield hit, the supplier has more power than it's using and is leaving money on the table out of caution. If the customer can swap easily, it isn't latent power, it's the correct price. Distinguishing the two is the whole judgment, and it's easier from inside the qualification process than from a spreadsheet.

What flips it: consolidation of the supplier's position or diversification of its customer base. The day it no longer depends on one buyer's goodwill, the pricing follows.

4. Optics and regulatory restraint

The rung: the most visible chokepoints. A supplier that raises prices too obviously during a shortage invites scrutiny, from customers, from governments, from the press. Some memory and materials players price a notch below what the market would bear precisely to avoid becoming the story.

The operator's read: this is real and it's a ceiling that can lift quietly. The restraint is about visibility, not capability, so the power gets captured in less legible ways, mix shift, allocation to higher-margin customers, quiet surcharges, rather than a headline price hike. The screen watching list price misses margin that's being taken through the back door.

What flips it: cover. When the whole industry is raising prices, no single name is the villain, and the restraint relaxes for everyone at once.

5. The business model hasn't caught up to the scarcity

The rung: the IP and tooling layer. The chip-design software duopoly and the core IP licensors sit on genuine chokepoints, complexity that only grows, and still often price on legacy per-seat or per-license models set before AI made their position load-bearing. Selling a scarce, essential thing the old way leaves the scarcity uncaptured.

The operator's read: this is the most valuable and slowest-moving form of latent power, because the fix isn't a price increase, it's a model change, toward value-based pricing, capacity reservation, or tolling. The same logic applies at the physical layer: packaging and firm power are starting to be sold as reserved capacity rather than per-unit, and whoever reprices the model, not just the sticker, captures far more than a percentage bump.

What flips it: a model change, which takes longer than a price change but reprices the whole revenue base when it lands.

The operator's edge, stated plainly

Every one of these is invisible to a screen for the same reason: the screen sees the price and treats it as the truth about the business. The operator sees the price and asks why it's set there, and the answer, trauma, a contract, a nervous customer, a fear of scrutiny, a stale model, tells you both that the power exists and roughly when it releases. That's the edge, and it's not a data edge. It's a context edge, the kind you only get from having been in the room.

The trap, named

There is one failure mode that turns this whole exercise into wishful thinking, and it has to be guarded against: mistaking absent power for latent power. Plenty of companies aren't charging more because they can't. The test is whether a genuine chokepoint exists underneath, real concentration, a switching cost, a qualification moat, a physical scarcity. If it does, the unused pricing is latent and the question is only when. If it doesn't, "not capitalizing yet" is just a story you're telling yourself about a commodity. Run every candidate through that gate before you believe the optionality is there.

Why it's the forward map

Mapping who already has pricing power tells you who won the last quarter. Mapping who holds it unused, and what flips each one, tells you where the next repricing comes from before it shows up in a print. The scar-tissue names reprice on belief, the contract names on the calendar, the customer-fearful names on diversification, the visible names on cover, the model-laggards on a model change. Those are five different clocks, and knowing which one governs a given chokepoint is the difference between guessing and timing.

You do not remove a bottleneck. You relocate it. And the bottlenecks that haven't yet charged what they're worth are not the weak links in this buildout. They are the strong ones that haven't yet decided to act like it. The moment they do is the moment the price you see today stops being the price.