Micron and SK hynix Selloffs: Test the Shared Nvidia Explanation

Key findings

• An equal-weight Micron and SK Hynix basket has a 22.75% average mapped Nvidia revenue-exposure indicator, with SK Hynix contributing 61.3% of the two percentages' simple sum.
• Nvidia-originated product or purchasing news has a named path to both suppliers, while a factory or execution problem confined to one supplier should not be propagated to the other without a shared-node mechanism.

A Micron–SK hynix selloff needs an event source before it gets a network explanation.

Altsets maps Nvidia as a shared customer of Micron and SK Hynix, but the two suppliers are not equally concentrated. Nvidia represents an estimated 27.88% of SK Hynix revenue and 17.62% of Micron revenue. When new information originates at Nvidia, those edges define a testable common transmission path. When the information originates inside only one memory supplier, they do not.

The shared-customer cluster is measurable

Across an equal-weight basket of the two suppliers, the average mapped Nvidia revenue exposure is 22.75%. This is a concentration indicator for comparing the basket with another supplier basket; it is not a predicted revenue loss or stock beta. Within the simple sum of the two revenue percentages, SK Hynix contributes 61.3% and Micron 38.7%. The shares show which supplier dominates the pair's mapped customer concentration, not how a portfolio will trade.

Public product announcements confirm that the shared node is commercially relevant to current products. Micron said its HBM4 was in high-volume production for Nvidia Vera Rubin. SK Hynix's multi-year Nvidia partnership covers Vera Rubin systems and several other Nvidia platforms. The relationship estimates measure economic importance; the announcements identify a current product and technology mechanism through which Nvidia-specific information can reach each supplier.

Event origin determines whether the network explanation fits

A Nvidia platform delay, purchasing change, or memory-specification change enters at the common customer. Both supplier edges then deserve review because both have a named business path to the event. Altsets ranks the starting exposure: SK Hynix's mapped Nvidia revenue share is 10.26 percentage points higher, or 58.2% higher relative to Micron's estimate.

A Micron fabrication, yield, or execution problem enters at Micron. The Nvidia to SK Hynix edge provides no mechanism for automatically propagating that company-specific problem to SK Hynix. The same rule applies in reverse to an issue confined to SK Hynix. A common end market is not enough; the event must alter the shared customer's demand or another node connected to both suppliers.

This distinction prevents two opposite errors. Treating every memory-stock decline as a Nvidia event overstates network contagion. Treating Nvidia-originated product news as relevant to only the supplier mentioned in one headline understates the common customer path.

A four-part evidence test separates the cases

First, locate the event's source. A supplier announcement, customer roadmap change, industry price move, and broad market shock begin at different nodes.

Second, require a relationship path. The Micron and SK Hynix edges pass this test for Nvidia-specific demand news because both suppliers have quantified customer exposure and current public product evidence. They do not pass it for a problem unique to one supplier's factory.

Third, compare directional concentration. For a shared Nvidia event, the 27.88% and 17.62% supplier-revenue estimates are the relevant figures. The customer-cost percentages answer a different question about Nvidia's sourcing mix and should not be averaged with supplier-revenue exposure.

Fourth, identify an event-specific mechanism. A product delay could change shipment timing; a specification change could alter qualification or product mix; a purchasing reduction could affect supplier volumes. If the mechanism cannot be stated, the observed price correlation is not yet supply chain evidence.

Different price moves do not disprove a shared event

Even with the same customer, Micron and SK Hynix can respond differently because the network measures company-wide commercial concentration, not product-level margin, contract protection, inventory timing, capacity availability, or investor expectations. SK Hynix's larger mapped percentage establishes greater customer concentration, but it does not establish a larger share-price response.

The converse also holds. A simultaneous selloff is not proof of a network event. Interest rates, currency moves, sector de-risking, or index flows can move both securities without changing the Nvidia relationship. Price co-movement is an observation to explain; the named edges and event mechanism are the evidence required for a supply chain diagnosis.

A residual event test can distinguish the explanations

For a Nvidia-originated announcement, estimate Micron and SK hynix returns relative to their local market and semiconductor factors over trading sessions that actually follow the release. Compare the supplier residuals with other memory companies whose Nvidia product exposure differs. The pre-event customer shares can enter as economic weights, but the two named suppliers alone are insufficient for a credible cross-sectional estimate.

A common residual response is consistent with the customer channel; a response across all semiconductor stocks favors a wider factor explanation. A supplier-only operating event is a separate sample. This is a proposed diagnostic, not evidence that a dated selloff has been causally attributed.

Sources

• Altsets supply chain dataset
• Micron HBM4 for Nvidia Vera Rubin
• SK Hynix and Nvidia multi-year technology partnership

Related research

• Do Supply Chain Links Improve Tail-Risk Models?

• Does Customer Volatility Show Up in Supplier Options?

• Is It a Supply Chain Spillover or Just Industry Momentum?

• Supply Chain Data for Forecast Uncertainty, Not Just Point Forecasts

Methodology

Read the methodology for this research.

Cite this research

Altsets Research. "Micron and SK hynix Selloffs: Test the Shared Nvidia Explanation." Published September 17, 2026. https://www.altsets.com/research/company-specific-vs-network-wide-selloff

Copy citationCanonical URLMethodology