AI Memory Stocks: Own the Shipment Cycle or the Equipment Cycle?

Key findings

• Nvidia accounts for an estimated 27.88% of SK Hynix revenue and 17.62% of Micron revenue, making SK Hynix's mapped direct customer concentration 10.26 percentage points, or 58.2%, higher.
• KLA's 4.59% Micron revenue exposure is a second-order equipment path: it is 6.07 times smaller than SK Hynix's Nvidia revenue percentage and depends on an intervening Micron manufacturing-investment decision.

Which AI-memory exposure matches the investment thesis: memory shipments or the equipment needed for new capacity?

Altsets maps three distinct positions in the AI-memory chain: SK Hynix and Micron supply Nvidia, while KLA supplies Micron. The relationships carry different kinds of exposure even though all three companies sit inside the same broad investment theme.

The first two edges measure direct exposure to Nvidia purchasing. The third is a second-order path: Nvidia demand can influence Micron's production and investment decisions, which can then influence KLA. Treating the three stocks as interchangeable expressions of "AI" would discard that structure.

The direct memory positions have materially different customer concentration

Nvidia's attributed share of SK Hynix revenue is 10.26 percentage points above its share of Micron revenue. Relative to Micron's 17.62%, SK Hynix's 27.88% exposure is 58.2% higher. That comparison uses the same metric on the same side of two supplier-customer relationships, so it is a valid concentration comparison. It is not a forecast of revenue growth or stock performance.

Public product evidence connects both edges to the same platform cycle. Micron said in March 2026 that it was in high-volume production of HBM4 designed for Nvidia Vera Rubin. SK Hynix's announced multi-year technology partnership with Nvidia covers memory for Vera Rubin systems as well as Vera CPUs, RTX Spark PCs, and Jetson Thor. The public announcements identify products and development scope; the Altsets percentages supply the company-level economic scale that the announcements do not disclose.

The combined evidence produces a specific distinction. SK Hynix has the larger mapped Nvidia revenue concentration, while its public partnership also spans more than one Nvidia platform. Micron has a smaller, though still material, mapped concentration and a product-specific HBM4 connection to Vera Rubin. A change in Nvidia demand belongs in both research files, but the network estimate gives SK Hynix the larger direct concentration to investigate first.

KLA is exposed through a different decision

KLA's relationship is not an alternative measurement of the same Nvidia edge. Altsets attributes 4.59% of KLA revenue to Micron and 2.84% of Micron's cost base to KLA. The immediate economic node is Micron, not Nvidia.

Comparing supplier-side revenue percentages shows how much more concentrated the direct customer paths are. SK Hynix's Nvidia percentage is 6.07 times KLA's Micron percentage, and Micron's Nvidia percentage is 3.84 times KLA's Micron percentage. Those ratios compare customer concentration at the respective suppliers. They do not say that one stock has 6.07 times the return sensitivity of another.

The transmission mechanisms also differ. Nvidia purchasing can reach SK Hynix and Micron through product volumes, mix, qualification, and contract terms. Nvidia demand reaches KLA only if it changes Micron's need for process-control equipment, whether through new capacity, technology transitions, or spending at specific facilities. Strong memory shipments can coexist with weak equipment orders if Micron uses existing capacity; equipment spending can also precede the revenue it is intended to support.

One theme therefore contains two separate clocks

The direct supplier edges operate on the memory-product cycle. The KLA edge operates on Micron's manufacturing-investment cycle. That timing difference is more useful than a generic claim that equipment is "less direct."

For a Nvidia platform event, the first comparison is SK Hynix's 27.88% and Micron's 17.62% attributed revenue exposure, followed by the product scope named in their announcements. For a Micron capacity or process-transition event, KLA's 4.59% customer concentration becomes the relevant edge. An event at Nvidia should not be propagated automatically to KLA without evidence that Micron's equipment demand changes.

What the comparison establishes, and what it leaves open

The Altsets estimates are company-wide relationship measures, not HBM-only revenue shares. The public announcements validate current product or partnership links but do not reconcile directly to the relationship percentages. KLA's second-order exposure cannot be calculated by multiplying 17.62% by 4.59% because the percentages have different denominators and no disclosed pass-through rate.

The defensible finding is narrower: the same AI-memory theme contains two direct Nvidia supplier edges with meaningfully different customer concentration and one equipment edge whose outcome depends on an intervening Micron investment decision. The network identifies which company decision must occur before each position benefits.

The stock-winners-from-shock guide examines how value can shift across positions in the same network after a disruption.

Sources

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

Related research

• HBM4’s Second Supply Chain: Logic Base Dies and Package Readiness

• Supply Chain Networks Shrink Your Effective Sample Size

• What Supply Chain Network Motifs Capture That Centrality Misses

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

Methodology

Read the methodology for this research.

Cite this research

Altsets Research. "AI Memory Stocks: Own the Shipment Cycle or the Equipment Cycle?." Published September 14, 2026. https://www.altsets.com/research/where-to-own-investment-theme

Copy citationCanonical URLMethodology