Shin-Etsu Chemical does $14 billion a year. They are the largest silicon wafer manufacturer on earth.

For one specific kind of wafer, they pay royalties to a 2,000-person company in the French Alps.

The French company is Soitec. Ticker SOI on Euronext Paris. We added it to the FCF Tracker on Monday. I do not own it personally yet. I want you to see how we decided whether it deserved the spot.

What Soitec actually makes, in a sentence you can repeat at dinner.

A normal silicon wafer is a slice of pure crystal, a few inches across, polished to mirror finish. Soitec takes one of those slices and bonds an atom-thin insulating layer inside it, like sliding a sheet of glass between two pieces of paper and welding them flat. The insulating layer stops electricity from leaking sideways. That sounds small. It cuts your phone's 5G chip power use by roughly a third. It lets AI accelerators switch faster without melting. It lets EV power chips handle 800 volts without frying.

The process is called Smart Cut. Soitec invented it in 1992 in a lab outside Grenoble, and they have been refining it for thirty-three years. There are 4,300+ patents around it. Shin-Etsu, the giant, decided licensing them was cheaper than designing around them. That is the entire story.

The moat, with the receipts.

Soitec holds 4,300+ patents on the SOI process. Shin-Etsu and SUMCO, the two largest wafer companies in the world, both pay Soitec to use it. When the giants in your category pay you instead of fighting you, that is the most defensible position a small company can hold.

Once a chip is qualified on a Soitec wafer, the customer cannot switch suppliers without 12-18 months of requalification and millions of dollars in engineering cost. STMicroelectronics, Samsung Foundry, GlobalFoundries, Tower Semiconductor are all on multi-year supply agreements. Switching costs are the closest thing to a moat in industrial businesses, and Soitec has them on every major customer.

A new SOI wafer fab costs more than €1 billion and five years to build. A competitor would have to commit that capital before they had a single qualified customer. Nobody is doing that. The moat compounds while everyone else looks at AI chips instead of the wafers underneath them.

58% of buyers now start their research in ChatGPT or Gemini, not Google. Most startups aren't showing up there yet.

The ones that are get cited by the AI tools their buyers, investors, and future hires already use. And they convert at 3×.

Download the free AEO Playbook for Startups from HubSpot and get the exact steps to start showing up. Five minutes to read.

Get the free playbook

The AI angle nobody is pricing.

The bottleneck inside an AI datacenter is no longer compute. It is the cable between the racks. Copper runs out of headroom around 200 gigabits per second. The industry is moving to optics, where data travels as light over fiber inside the datacenter itself. The shorthand is "co-packaged optics."

Optical chips need a special kind of wafer called Photonics-SOI. There are three companies in the world that can make it at production scale. Soitec is one of them.

Nvidia and Broadcom have published 2026-2028 roadmaps that depend on this transition happening. The volumes are small today. The volumes structurally have to be large in three years. You do not need to predict the timing. You need to know which suppliers it has to flow through.

The numbers, no rounding.

What would have to go wrong.

I argue against every name we add. Here is the bear case for Soitec, in plain terms.

SOI demand tracks smartphone units, automotive production, and 5G build-out. All three can soften at once, and 2023 proved it. If you cannot sit through a 30% drawdown without selling, you are in the wrong name.

If Apple decides the next iPhone RF design moves off SOI, the revenue hit is real and immediate. The current pipeline says they are not. Pipelines lie.

The photonics revenue ramp depends on hyperscaler timelines. If co-packaged optics slips from 2027 to 2029, the upside thesis pushes out by years. You are paid to wait, not to be right on the calendar.

China is pushing a domestic SOI alternative through SMIC and others. The quality is not there yet. The quality may eventually be there.

If you held this through all of that, would the moat still be intact in 2030? My honest answer is yes. That is why it goes on the watchlist instead of in the trash.

How we are going to size it (and how you would).

The Compounding Portfolio holds 8-9 names. Each one is sized so a 50% drawdown in any single position does not break the whole. Soitec, if it gets added, sits in the 4-7% allocation band. Defensible quality, not portfolio-defining conviction. The kind of name you accumulate over 12-18 months in tranches, never in one buy.

Three triggers would move it from watchlist to real money:

Until at least one of those hits, we watch.

The frame you can reuse.

Forget Soitec for a second. The thing worth taking from this email is the question we asked.

When a small specialist company supplies a critical piece of a much larger industry, ask: "Do the giants pay them, or fight them?"

If the giants pay them, the moat is real. If the giants are trying to displace them, the clock is ticking. That single question separates real industrial moats from companies that look like moats until the buyer decides to in-house production.

Try it on the next industrial name in your watchlist. It will tell you in five minutes whether to keep researching.

— Simon

Get the FCF Tracker spreadsheet.. the same one we just added Soitec, inside the Sprint Club → strategysprints.com

Also worth watching: The Sales Show — B2B sales acceleration for founders. 800+ episodes.
Subscribe → https://thesalesshow.beehiiv.com

Disclosure: We do not currently hold a position in SOI as of publication. Soitec is on the FCF Tracker watchlist, not in the Compounding Portfolio. Nothing in this email is investment advice. Do your own work.