The HBM Suppliers Beneath SK hynix and Samsung

Research question: Which suppliers below SK hynix and Samsung can have disproportionately large financial exposure to HBM-related spending?

HBM suppliers by process layer

The table answers the literal supplier question without implying that every company participates in every HBM stack. It lists only companies documented in this article and separates disclosed orders, mapped relationships, and broader product context.

This is a durable way to use the page: start with the layer a HBM question concerns, then read the supporting evidence and its boundary. The page is not a complete global supplier directory.

The U.S. HBM manufacturing analysis adds the geographic question: which parts of the memory, materials, logic, packaging, and integration chain remain dependent on Korea, Japan, and Taiwan even as U.S. capacity expands?

HBM4 turns memory capex into supplier concentration

Hanmi Semiconductor is the clearest current example.

In June, Hanmi disclosed a 44.2 billion won order from SK hynix for its TC BONDER 4.5 GRIFFIN equipment specifically for HBM4 manufacturing. The contract alone was equal to 7.66% of Hanmi's entire 2025 revenue. [3]

That is the kind of number that gets obscured when HBM analysis stops at the memory manufacturer.

Thermo-compression bonding is part of the process used to build vertically stacked HBM packages. For SK hynix, buying another group of bonders is one component of a much larger production system. For Hanmi, a single HBM4 equipment order can equal nearly one-twelfth of the previous year's revenue.

The relationship therefore has a different economic meaning depending on which side an investor owns.

Hanmi's exposure can rise sharply when SK hynix accelerates HBM capacity. The reverse conclusion does not automatically follow. The public contract disclosure tells us how material this particular order is to Hanmi, but it does not provide an Altsets Customer Cost % for the Hanmi to SK hynix relationship. It would therefore be wrong to turn the 7.66% figure into a claim about SK hynix's cost dependence or Hanmi's technical irreplaceability.

What it does establish is operating leverage to customer spending. A comparatively small change in SK hynix's equipment purchasing can translate into a much larger change in the revenue opportunity available to a specialized supplier.

That is the first layer investors should look for below HBM.

Samsung shows the same asymmetry at the materials layer

The pattern is not limited to packaging equipment.

Altsets maps Shin-Etsu Chemical as a supplier to Samsung Electronics. Samsung represents 2.43% of Shin-Etsu's revenue, while Shin-Etsu represents just 0.29% of Samsung's COGS.

Those metrics measure different sides of the same relationship. Supplier Revenue % measures Samsung's importance to Shin-Etsu's revenue. Customer Cost % measures Shin-Etsu's importance within Samsung's cost base.

The resulting asymmetry is substantial. Samsung's economic importance to Shin-Etsu is more than eight times Shin-Etsu's measured share of Samsung's COGS.

This does not mean 2.43% of Shin-Etsu's revenue comes from HBM. Altsets relationships are company-level relationships, and Samsung operates businesses far beyond HBM. It also does not prove that Samsung could easily replace Shin-Etsu. Technical qualification and economic dependence are separate questions.

But the relationship illustrates why the HBM supply chain should not be analyzed only by searching for the component with the largest cost share.

A supplier can be strategically relevant to semiconductor manufacturing while remaining almost immaterial to the customer's consolidated cost structure. At the same time, the customer can be meaningful enough to the supplier that changes in purchasing, qualification, capacity plans, or technology transitions matter disproportionately to the supplier's financial results.

HBM4 can intensify exactly this type of asymmetry.

As SK hynix and Samsung move toward more advanced DRAM processes, higher stack counts, tighter packaging tolerances, and eventually newer bonding architectures, they create demand across front-end equipment, materials, inspection, bonding, and packaging. Yet the memory producers are enormous diversified buyers. Many companies selling into those expansions are not.

That makes customer concentration one of the more useful ways to search for second-order HBM exposure.

An HBM supplier whose sales rise by hundreds of millions of dollars may barely change Samsung's or SK hynix's total manufacturing economics. The same spending can transform the revenue mix of a smaller equipment company.

The investment signal is asymmetry, not simply scarcity

The most interesting companies below SK hynix and Samsung are therefore not necessarily the suppliers with the largest absolute contracts.

They are companies for which HBM-related purchasing is large relative to the supplier's own revenue base.

Hanmi's June HBM4 order is unusually clean evidence because the contract is explicitly tied to HBM4 and immediately represents 7.66% of prior-year revenue. Shin-Etsu provides a broader company-level example of the same structure: Samsung is materially more important to the supplier's revenue than the supplier is to Samsung's costs.

Those are different relationships, and neither proves technical dependence. But both point to the same investment mechanism.

When HBM capacity expands, the percentage change in economic exposure can be much larger below the memory manufacturer than at the memory manufacturer itself.

For investors searching beyond SK hynix and Samsung, that is the part of the HBM supply chain worth measuring: not simply who supplies HBM production, but whose financial results can become unusually dependent on the companies producing it.

HBM4 is turning SK hynix and Samsung into increasingly important demand centers for specialized semiconductor suppliers. The more asymmetric exposure can sit downstream from their capital budgets: Hanmi received a single HBM4 equipment order equal to 7.66% of annual revenue, while Altsets shows Samsung accounting for 2.43% of Shin-Etsu revenue even though Shin-Etsu represents only 0.29% of Samsung COGS.

The key distinction is economic dependence. A supplier does not have to be a large cost item for SK hynix or Samsung to have significant exposure to their HBM expansion. In many cases, that imbalance is precisely what makes the second layer of the HBM supply chain more sensitive to the cycle than the giants at the center of it.

An HBM equipment order and a materials relationship need different clocks

The disclosed Hanmi order equal to 7.66% of prior-year revenue is a product-specific contract comparison. Shin-Etsu's 2.43% Samsung revenue share is a company-level relationship estimate. Ranking 7.66% against 2.43% as if both were annual HBM sales would mix a contract award with a relationship snapshot.

For Hanmi, the decisive next observations are shipment, acceptance and revenue recognition for the named order. For Shin-Etsu, they are the product portion and recurring purchasing behind the Samsung edge. A contract win can create a lumpy equipment cycle, while recurring materials can track ongoing production. The supplier screen should preserve that distinction even when both routes originate in the same memory expansion.

Sources

1. "SK hynix Announces 2Q26 Financial Results," SK hynix Newsroom, August 2026. https://news.skhynix.com/en/q2-2026-business-results/
2. "Samsung Ships Industry-First Commercial HBM4 With Ultimate Performance for AI Computing," Samsung Global Newsroom, February 12, 2026. https://news.samsung.com/global/samsung-ships-industry-first-commercial-hbm4-with-ultimate-performance-for-ai-computing
3. "HBM4 manufacturing TC BONDER 4.5 GRIFFIN equipment order," Korea Exchange disclosure by Hanmi Semiconductor, June 8, 2026. https://kind.krx.co.kr/external/2026/06/08/000436/20260520000503/91370.htm

This supplier layer belongs inside the geographic and customer context described in U.S. HBM manufacturing's Korean, Japanese, and Taiwanese dependencies. The customer comparison in SK hynix versus Micron's Nvidia exposure then shows why the same HBM cycle can matter differently across memory companies.

Go Deeper