How to Build a Relative-Value Screen Around a Shared Customer
August 25, 2026
Altsets
Research by Altsets Research
Use common customer exposure to choose economically comparable suppliers, then investigate why their fundamentals or prices diverge without using relationship percentages as trading hedge ratios.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- SK Hynix and Micron share Nvidia as a quantified customer at 27.88% and 17.62% of supplier revenue, creating a common economic factor for relative-value research.
- The exposure percentages describe customer concentration, not stock beta or hedge ratios, so product mix, currency, valuation, liquidity, and reporting timing still need separate controls.
Build a shared-customer relative-value screen by finding suppliers exposed to the same customer, comparing that customer's share of each supplier's business, and then researching why their fundamentals or prices might diverge. The exposure percentages make the pair economically coherent; they are not a trading hedge ratio. SK Hynix and Micron are both mapped as Nvidia suppliers, with Nvidia representing 27.88% of SK Hynix revenue and 17.62% of Micron revenue in the displayed data.
The relationship data does not answer the trade. It makes the comparison economically coherent.
Why a shared customer is useful for pair selection
Traditional pair screens often start with sector or correlation. Two companies can both be classified as memory manufacturers while having very different customer exposure. A shared-customer screen adds another condition: Both companies should be tied to the same identifiable demand source.
That makes it easier to separate common demand from company-specific differences. In this example, Nvidia is a common mapped customer. The supplier-revenue percentages show that Nvidia matters to both suppliers, but more heavily to SK Hynix in the displayed data.
The pair is similar enough to compare, but not identical
Public company information shows both companies are exposed to AI memory. SK Hynix said HBM4 mass shipments began in Q2 2026 and announced a multi-year Nvidia partnership covering next-generation AI memory. Micron said HBM4 was in high-volume shipments for its lead customer's platform and described strong AI-driven data-center memory demand.
That gives the pair a common economic theme. But they differ in customer concentration, product mix, geography, capital intensity, NAND exposure, contract structure, currency, balance sheet, valuation, and reporting calendar. Those differences are exactly what relative-value research should investigate.
Do not call the exposure difference a hedge ratio
A tempting shortcut would be to divide 27.88 by 17.62 and use the result as a trade hedge ratio. That would be unjustified. Supplier revenue percentages are not stock betas.
They do not measure price sensitivity. They do not control for margins, market capitalization, volatility, foreign exchange, product mix, or operating leverage. The ratio can describe customer concentration. It cannot size a long-short position by itself.
Use the shared customer to define the common factor
The better use is conceptual. Treat Nvidia-linked AI demand as one common factor. Then ask what remains different between the two suppliers. For example:
- Does one have stronger HBM execution?
- Does one have more exposure to general DRAM or NAND?
- Does one have tighter long-term customer commitments?
- Are capacity additions different?
- Are margins moving differently?
- Is one valuation pricing in more of the same demand story?
The relationship graph tells you that these questions belong in the same comparison.
Public announcements can strengthen the common-factor case
SK Hynix's June 2026 announcement directly named Nvidia in a multi-year memory partnership. Micron's public HBM4 statement did not identify the lead customer. That creates an evidence asymmetry.
For SK Hynix, Nvidia-specific commercial context is publicly confirmed. For Micron, the proprietary relationship establishes the mapped Nvidia edge while public commentary confirms strong HBM4 and AI-memory demand more broadly. A relative-value researcher should preserve that difference in evidence quality.
Reporting dates create another dimension
The pair can also be useful around earnings. If Nvidia reports first, both suppliers become read-through candidates. If one memory supplier reports before the other, the first supplier's commentary can become another input into the second.
This creates several possible comparison windows without requiring a permanent long-short position. The options-trading workflow covers how event timing can narrow a watchlist. Relative-value research adds the question of why two economically linked suppliers may respond differently to the same demand signal.
What should be controlled before testing the pair
A serious quantitative or discretionary comparison should control for market beta, currency, country risk, index membership, market capitalization, liquidity, memory product mix, semiconductor cycle exposure, reporting dates, and major company-specific events. Otherwise a divergence attributed to Nvidia exposure may actually come from an unrelated factor. Supply-chain data improves pair selection. It does not remove the need for factor control.
A repeatable shared-customer pair workflow
- Choose a customer or demand driver.
- Find multiple suppliers with quantified exposure to that customer.
- Require enough economic exposure for the common factor to matter.
- Compare supplier-revenue percentages.
- Validate product overlap with public company information.
- Identify company-specific differences.
- Define the hypothesis before looking at returns.
- Control for market, currency, and sector factors.
- Use point-in-time relationships for any historical test.
- Keep exposure weights separate from trading hedge ratios.
The economic-basket guide explains how to group several suppliers around one customer. This article narrows the idea to a two-company relative-value research design, where the goal is to explain divergence rather than simply own the shared theme. For point-in-time testing rules, read the look-ahead-bias guide. For relationship methodology, read the Altsets supply-chain data methodology.