2026.02.10 2026.02.10

Disco (6146): The “Hidden Ruler” Behind NVIDIA with 80% Global Market Share

s7たち

This article may contain affiliate links. A small commission may be earned at no additional cost to you.

Disco (6146): The “Hidden Ruler” Behind NVIDIA with 80% Global Market Share

💡 Executive Summary

The Monopoly: Disco holds a staggering 70-80% global market share in semiconductor “cutting and grinding” equipment. It is a “Global Niche Top” company.
The HBM Gatekeeper: AI chips (like NVIDIA’s H100/Blackwell) require HBM (High Bandwidth Memory). HBM cannot be manufactured without Disco’s ultra-thin grinding technology.
The Cash Machine: They operate on a “Razor and Blade” business model. They sell the machines, but they make massive profits selling the disposable diamond blades required to run them.

1. Introduction: The “Achilles’ Heel” of NVIDIA

If this one Japanese company stopped production tomorrow, shipments of NVIDIA GPUs would halt, and the global AI revolution would hit a brick wall.

The company is Disco Corporation (6146).

Most investors focus on the giants: ASML (who draws the circuits) and TSMC (who bakes the chips). But in the “Back-end” (packaging and finishing) process, Disco wields god-like power.

Why is this quiet Japanese manufacturer called the “Shadow Ruler” of the AI Gold Rush? Let’s look at the numbers.

2. What Do They Actually Do? (The Pizza Analogy)

Semiconductor manufacturing is complex. Let’s simplify it using a Pizza Delivery analogy.

Process	Role (The Pizza Analogy)	Key Players
Front-end	“Baking & Toppings” Creating the complex circuitry on the silicon wafer (the dough).	ASML (Lithography) Tokyo Electron (Coating)
Back-end	“Slicing & Boxing” Cutting the wafer into individual chips, stacking them, and packaging them.	Disco (Dicing/Grinding) Advantest (Testing)

The Investor Insight:

Historically, the “Front-end” (ASML) was the star. But in the AI era, the value has shifted to the “Back-end.”

Why? Because modern AI chips use 3D Stacking (Chiplets/HBM). You cannot stack chips unless you cut and grind them with atomic precision. That is Disco’s territory.

3. Why is Disco Essential for HBM (AI Memory)?

The biggest bottleneck in AI hardware right now is HBM (High Bandwidth Memory).

HBM works by stacking memory chips vertically, like a skyscraper (8, 12, or even 16 floors high).

The “Grinding” Bottleneck

To stack 12 chips into a tiny package, you must shave each silicon chip down to be thinner than a sheet of paper.

Standard Wafer Thickness: ~775μm (microns)
Target Thickness for HBM: ~50μm (microns)

Disco’s “Grinders” shave the silicon down to 50μm without cracking it.

If the grinder vibrates even slightly, the wafer cracks, and $50,000 worth of chips are ruined.

Only Disco can mass-produce machines with this level of stability. As HBM demand grows, demand for Disco’s high-end grinders grows exponentially.

4. The Business Model: The Ultimate “Razor & Blade”

Disco’s technology is impressive, but their Business Model is what makes the stock a long-term winner.

The Razor (The Machine): Precision Dicing Saws and Grinders that cost millions of dollars.
The Blade (The Consumable): To cut diamond-hard silicon, you need Diamond Blades and Grinding Wheels.

These blades wear out constantly.

As long as TSMC’s factories are running, they must buy new blades from Disco every single day.

This Recurring Revenue model creates a massive safety net. Even if semiconductor capital spending slows down (machines sell less), the factories still need blades to run existing machines. This keeps Disco’s profit margins exceptionally high (30-40%).

5. Risks & Competitors

To be fair, let’s look at the potential downsides.

Currency Risk (FX): A significant portion of sales are overseas. A sudden, sharp appreciation of the Japanese Yen could hurt reported earnings.
Consumer Tech Slowdown: While AI (HBM) is booming, standard chips for Smartphones and PCs make up a large volume. If global consumer spending drops, the demand for standard “blades” may decrease.