The Companies Two Layers Beneath NVIDIA

The concentration gets larger as the companies get smaller

The direct memory relationships are already unusually material. Altsets estimates that SK hynix represents 27.33% of NVIDIA's COGS while NVIDIA represents 27.88% of SK hynix revenue. Micron represents 14.00% of NVIDIA's COGS, while NVIDIA represents 17.62% of Micron revenue.

Those figures describe substantial economic relationships in both directions. They do not prove that every dollar is HBM-related, nor do they establish technical substitutability. But they make SK hynix and Micron useful starting points for finding companies further down the chain.

SIMMTECH is one of the clearest examples.

The Korean listed PCB manufacturer reported KRW 1.411 trillion of revenue in 2025. SK hynix accounted for KRW 230.4 billion, or 16% of revenue. Micron accounted for another KRW 197.0 billion, or 14%.[2]

Together, two major NVIDIA memory suppliers represented 30% of SIMMTECH's revenue.

That is a different type of NVIDIA exposure from owning a direct supplier. NVIDIA itself is not the disclosed customer. Instead, SIMMTECH has substantial customer concentration in two companies whose own economics are materially connected to NVIDIA.

The distinction is important. It would be incorrect to multiply these percentages together and call the result SIMMTECH's "NVIDIA exposure." SIMMTECH's sales to Micron and SK hynix are not disclosed according to the end customer those products eventually support.

The useful finding is the concentration chain itself. Nearly one-third of SIMMTECH's revenue comes from two memory manufacturers that both have unusually large NVIDIA relationships. If those manufacturers expand procurement, packaging capacity or related production infrastructure, SIMMTECH has a comparatively small revenue base over which that spending can land. If their investment cycle weakens, the same concentration can work in reverse.

SIMMTECH also shows why looking only for single-customer dependencies misses part of the story. SK hynix exposure fell from 21% of revenue in 2024 to 16% in 2025, while Micron remained significant at 14%.[2] The company is not simply a proxy for one memory producer. It is exposed to multiple participants in the same AI-memory investment cycle.

A second path runs through the fab itself

Japan's Nomura Micro Science sits in a different part of the chain.

Nomura specializes in ultrapure-water systems used in semiconductor manufacturing. In its fiscal year ended March 2026, the company generated JPY 56.25 billion of consolidated sales. SK hynix alone accounted for JPY 6.85 billion.[3]

That makes SK hynix approximately 12.2% of Nomura Micro Science's total revenue.

Unlike a component relationship, this path is tied to manufacturing infrastructure. Advanced semiconductor production requires extremely pure process water, so Nomura's exposure is connected to the construction, expansion and operation of fabrication capacity rather than to a specific NVIDIA GPU.

Again, the relationship should not be overstated. The disclosure does not establish that Nomura equipment serving SK hynix is dedicated to HBM, Rubin or NVIDIA production. What it does establish is economic dependence: a single memory manufacturer two layers beneath NVIDIA generated more than one-eighth of Nomura's annual sales.

That is precisely the type of asymmetry that becomes less visible when investors stop their supply-chain analysis at NVIDIA's direct suppliers.

The investment implication is dependency stacking

The common screen for AI infrastructure starts with NVIDIA, TSMC, SK hynix and Micron. The deeper screen asks which smaller companies depend economically on those firms.

SIMMTECH and Nomura Micro Science illustrate two different versions of that exposure. SIMMTECH has customer concentration across two major NVIDIA memory suppliers, with SK hynix and Micron contributing 30% of revenue combined. Nomura has a narrower infrastructure relationship, with SK hynix contributing about 12.2% of group sales.

Neither relationship proves that NVIDIA demand will translate directly into revenue growth, margins or stock performance. These companies have other customers, other products and their own operational cycles. But the concentration reveals where second-order sensitivity can accumulate.

The key point is not that every supplier of an NVIDIA supplier is an AI beneficiary. It is that economic dependence can become more concentrated as the supply chain gets smaller.

For NVIDIA, Micron and SK hynix are major vendors. For companies such as SIMMTECH and Nomura Micro Science, those same memory manufacturers can represent a much larger portion of the business than their position several layers away from the GPU would suggest.

As Rubin moves into production, the most interesting NVIDIA supply-chain exposures may therefore be the companies that never invoice NVIDIA at all.

Sources

1. NVIDIA Corp., "Q2 Fiscal 2027 Earnings Call," August 26, 2026, NVIDIA Investor Relations.

2. SIMMTECH Co., Ltd., "2025 Annual Business Report," March 17, 2026, Korea Exchange and Financial Supervisory Service disclosure.

3. Nomura Micro Science Co., Ltd., "Consolidated Financial Results for the Fiscal Year Ended March 31, 2026," May 14, 2026, Tokyo Stock Exchange disclosure.

How to Cite This

For research inquiries or data access: press@altsets.com

Go Deeper