SK ecoplant’s SK hynix Exposure Is a Fab-Spending Question

Key findings

• A hypothetical 10% reduction in SK ecoplant sales to SK hynix implies a 3.649% reduction in SK ecoplant total revenue, other customer sales fixed.
• Fab infrastructure and recurring materials can respond differently to memory demand; the company-level edge does not allocate the relationship between them.
• Multiplying the SK ecoplant revenue share by SK hynix’s Nvidia share does not produce SK ecoplant Nvidia revenue exposure.

SK ecoplant sits upstream of SK hynix, but calling it another Nvidia demand proxy skips the decision that matters. Does stronger memory demand require new fab infrastructure, increase recurring materials consumption, or use capacity that is already in place?

Altsets estimates SK hynix at 36.49% of SK ecoplant revenue in the July 30, 2026 snapshot. That makes the memory producer a substantial customer. It does not identify which spending channel accounts for the relationship.

A large customer share can conceal two different cycles

SK ecoplant's infrastructure and semiconductor materials descriptions cover different functions. Infrastructure is tied to project approval, construction and delivery. Recurring materials are tied more closely to manufacturing activity. The Altsets company-level estimate does not provide that split.

That distinction produces a more useful investment opinion than “second-order AI exposure.” Memory shipments can strengthen while a fab project is delayed. A fab can also be built before the memory sold from it generates revenue. The relationship should be monitored against SK hynix project and production decisions separately.

What the concentration permits us to calculate

If SK ecoplant sales to SK hynix fell 10%, total SK ecoplant revenue would fall 3.649%, assuming all other revenue stayed fixed. A 20% customer-sales decline would contribute −7.298%. These are sensitivities applied to the estimated starting share, not observed project cancellations.

The 18.81% customer-cost estimate describes the opposite direction, inside SK hynix's cost base. It does not establish that a disruption at SK ecoplant would remove 18.81% of SK hynix production. Nor is the $3.1 billion relationship estimate a disclosed construction backlog.

Why the Nvidia path needs an intervening variable

The network can run SK ecoplant → SK hynix → Nvidia. It is tempting to multiply customer revenue shares along that path and report a precise Nvidia exposure. That arithmetic would combine percentages of two different companies' revenue without establishing what portion of SK ecoplant work serves Nvidia-related output.

A more credible hypothesis uses an intervening variable: Nvidia demand affects SK hynix's manufacturing plans, which affect the relevant SK ecoplant project or recurring purchases. Historical project announcements, utilization and relationship changes can test each step. If SK hynix meets demand from existing capacity, the construction channel may be weak even when the downstream memory relationship grows.

Sources

• Altsets supply chain dataset
• Altsets quantified exposure data
• SK ecoplant semiconductor infrastructure services
• SK ecoplant semiconductor gases and materials

Related research

• SK hynix Has More Revenue Exposure to Nvidia Than Apple

• U.S. HBM Manufacturing Still Depends on Korea, Japan, and Taiwan

• Map Customer Facility Disasters to Supplier Capex and Cash Risk

• The External Catalysts That Belong on a Stock's Calendar

Methodology

Read the methodology for this research.

Cite this research

Altsets Research. "SK ecoplant’s SK hynix Exposure Is a Fab-Spending Question." Published August 6, 2026. https://www.altsets.com/research/sk-ecoplant-sk-hynix-upstream-dependency

Copy citationCanonical URLMethodology