Customer Concentration Should Widen the Revenue Scenario Range, Not Automatically Lower the Forecast
September 14, 2026
Altsets
Research by Altsets Research
A dominant customer increases the sensitivity of the revenue path to one outside company. That should widen upside and downside scenarios before it automatically changes the central forecast.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- The supplied SK Hynix data places Nvidia at 27.88% of revenue and Apple at 8.75%, creating greater single-customer scenario sensitivity than the displayed Shin-Etsu relationships with TSMC, Samsung, and Intel.
- The relationship percentages identify economic concentration but do not calibrate statistical forecast intervals, so they should guide scenario design rather than be converted directly into forecast error bands.
Customer concentration should usually widen the range of plausible revenue outcomes before it changes the central forecast. A supplier with one dominant customer can still have an excellent expected growth path, but more of that path depends on one outside company's purchasing decisions. A supplier with several meaningful customers can have a similar expected growth rate with less sensitivity to any single buyer.
SK Hynix has a narrower set of large displayed customer bets
The supplied Altsets data shows Nvidia representing 27.88% of SK Hynix revenue and Apple representing 8.75%. Those two displayed relationships alone show why customer-specific scenarios matter to the SK Hynix revenue outlook.
A strong Nvidia demand case can materially improve the supplier thesis, but a weak Nvidia purchasing case can also move the revenue scenario substantially. That does not mean the investor should lower the expected revenue estimate simply because the customer is large. It means the range around that estimate should reflect the fact that one company controls a meaningful portion of the supplier's demand path.
Apple adds another important customer, but the displayed Nvidia relationship is more than three times as large by supplier revenue percentage. The customer mix is therefore not symmetric.
Shin-Etsu shows a different displayed shape
The supplied Shin-Etsu Chemical data shows TSMC at 4.02% of supplier revenue, Samsung Electronics at 2.43%, and Intel at 1.79%. The three displayed semiconductor customers sum to 8.24%, with no single relationship approaching the supplier-side weight of Nvidia for SK Hynix.
That does not automatically make Shin-Etsu safer. The three customers still participate in the semiconductor cycle, and the supplied set is not the company's complete customer base. What it does show is a different form of customer-specific sensitivity. One TSMC-specific shock has a smaller displayed supplier-revenue weight than one Nvidia-specific shock has for SK Hynix.
The scenario design should therefore differ even if both companies are exposed to semiconductor demand.
A wider range is not the same as a bearish forecast
This distinction matters because concentration analysis is often turned too quickly into a negative stock opinion. A large strategic customer can support rapid growth, provide product validation, improve planning, and justify investment. Academic research has found both performance benefits and risk costs from customer concentration, including evidence of an inverted relationship between concentration and supplier financial performance.
The central forecast can remain strong if the major customer is healthy and the supplier is well positioned. The concentration becomes visible in the downside and upside scenarios. A supplier may have more upside if one major customer accelerates unexpectedly and more downside if that same customer cuts orders.
That is a different conclusion from saying concentration is simply bad.
The graph tells the investor which scenario deserves the most width
A revenue model does not need a separate scenario for every customer. The relationship data can prioritize the few outside companies whose supplier-revenue percentages are large enough to materially alter the result.
For SK Hynix, Nvidia clearly belongs in that group based on the supplied 27.88% relationship. Apple also matters at 8.75%. For Shin-Etsu, the displayed TSMC, Samsung, and Intel exposures are more distributed, suggesting that a single-customer scenario should be smaller relative to an industry-wide semiconductor scenario.
No percentage should be translated directly into a forecast error band. The percentages tell the investor where the forecast is economically concentrated. Historical forecast errors and actual customer behavior would be needed to calibrate a statistical interval.
The conclusion is to separate expected growth from concentration risk
Customer concentration should change how uncertain the revenue path is, not automatically whether the central forecast is high or low. The supplied data shows a more concentrated Nvidia dependence for SK Hynix and a more distributed displayed semiconductor customer set for Shin-Etsu. Those structures deserve different scenario ranges even if the investor expects both companies to grow.
The customer-mix sensitivity guide explains how customer weights can be used in first-order scenarios. The customer concentration thesis guide explains why a major customer can strengthen and weaken an investment case at the same time.
For relationship definitions and evidence limits, read the Altsets methodology.