Why Anonymous Customer Concentration Is Hard to Track Across Filings
August 14, 2026
Altsets
Research by Altsets Research
Explain why anonymous customer labels cannot be treated as permanent identities and why defensible historical analysis requires extensive cross-referencing across dated relationship and disclosure evidence.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- Nvidia's filings show that anonymous customer labels are period-specific reporting references rather than durable company identifiers, so percentage continuity cannot establish counterparty continuity.
- Defensible historical identification can require extensive cross-referencing across dated filings, relationship history, channel evidence, company disclosures, and proprietary records, while a matching percentage remains only a clue.
Nvidia's filings show why anonymous customer concentration is much harder to follow than a simple percentage chart suggests. For fiscal 2025, Nvidia disclosed three direct customers at 12%, 11%, and 11% of annual revenue. In the third quarter of fiscal 2026, it disclosed four direct customers above 10% of quarterly revenue at 22%, 15%, 13%, and 11%.
The filing also warns that customer references used in one period can represent different customers from those shown in another period. That single disclosure breaks the simplest possible tracking method.
Customer A is not a permanent company identifier.
A repeated label is not continuity evidence
An investor can easily build a chart that appears precise: Customer A: 12%, then 22%, then 21%. The problem is that the chart may be connecting different companies.
Anonymous labels are reporting devices inside a specific filing period. They are not stable identifiers like a legal-entity ID or security identifier. So a longitudinal customer-concentration analysis has to begin with uncertainty, not assumed continuity.
The percentages themselves are still useful
Even when customer identity is unresolved, the concentration pattern contains information. Across Nvidia's recent disclosures, major direct customers repeatedly account for double-digit shares of revenue. That tells the investor something real about the concentration structure.
What it does not tell the investor is which named company sits behind every percentage in every period. The distinction matters because concentration analysis and counterparty identification are separate research problems.
Different reporting windows create another layer of difficulty
Quarterly and annual percentages are not interchangeable. A customer can exceed 20% of revenue in one quarter and represent a smaller percentage across the full year. Another customer can move above or below the disclosure threshold as purchasing schedules change.
So any serious historical comparison has to preserve the original period. A quarterly concentration disclosure cannot be silently stitched into an annual series as though the denominator were unchanged.
Direct and indirect customers complicate the picture further
Nvidia distinguishes direct customers from indirect customers. A direct buyer can be an ODM, OEM, distributor, cloud provider, system integrator, or another channel participant. The economically important end user may therefore be different from the legal counterparty whose purchases appear in the direct-customer concentration disclosure.
This means even a correct company relationship can sit at the wrong level of the commercial chain for a particular filing statistic. The direct-versus-indirect customer guide explains why those layers cannot simply be collapsed.
Named relationship data does not solve identity by itself
The supplied Altsets Nvidia network contains named downstream relationships involving Quanta Computer, Samsung Electronics, Amazon, Microsoft, and Super Micro Computer. That is useful because it gives the researcher real counterparties and commercial paths to investigate. It does not prove that any one of those companies corresponds to Customer A, Customer B, or another anonymous filing label.
A named edge and an anonymous percentage are two pieces of evidence. They only become an identity claim when the surrounding evidence supports the connection.
Reliable identification requires far more cross-referencing than one filing
This is where the work becomes difficult. A defensible historical identification can require cross-referencing multiple dated sources, including:
- company filings;
- earnings materials;
- supplier and customer disclosures;
- historical relationship evidence;
- product and channel information;
- entity-name changes;
- reporting-period changes;
- public and proprietary records that may not use the same naming conventions.
The important point is not the exact mechanics of that work. The important point is that one percentage match, one press release, or one repeated anonymous label is not enough. That is why normalized relationship data is valuable in the first place.
Historical evidence has to exist at the historical date
Today's relationship map cannot automatically explain a filing from two years ago. A customer can enter or leave the network. A direct relationship can become indirect.
A supplier can change legal entities. A customer can cross a disclosure threshold only temporarily. Point-in-time relationship history therefore matters when historical concentration is being studied. The point-in-time backtesting guide explains why current relationships should not be projected backward without dated evidence.
A percentage match is especially dangerous
Suppose an anonymous filing customer represents 14% of revenue. Suppose a named relationship estimate also happens to be near 14%. That similarity is not proof of identity. The values can differ in:
- reporting date;
- methodology;
- denominator;
- direct versus indirect channel level;
- accounting treatment;
- estimate construction.
Matching numbers can be a clue. They are not a conclusion.
What an investor can safely learn without identifying the customer
Even when the name remains unresolved, the filing still supports several useful observations: how many customers exceed a disclosure threshold, how large the biggest disclosed concentration is, whether concentration appears to be rising or falling, whether direct and indirect channels are becoming more important, whether receivable concentration differs from revenue concentration, and whether management changes the way it describes the customer base. That is enough to make the filing analytically useful without forcing a false identity.
Why this matters for investment research
False customer identification can contaminate several downstream analyses. It can create a fake demand-read-through thesis. It can produce an incorrect event-exposure map.
It can make a portfolio appear more concentrated than it is. It can cause a researcher to assign public news to the wrong counterparty. The cost of being wrong is therefore larger than one mislabeled chart.
Altsets should reduce the research burden, not pretend the problem is easy
The value of a supply-chain relationship dataset is not that anonymous customer identification becomes trivial. It is that a large amount of fragmented relationship evidence can be organized into a more usable economic network. The difficult work still requires evidence discipline.
The result should make serious cross-company research more practical without implying that every anonymous disclosure can be decoded from one metric. The unnamed-major-customer guide discusses what named relationship data can and cannot contribute to that research. For relationship definitions and historical-data limits, read the Altsets methodology. Browse Supply-Chain Data Use Cases for other filing and relationship-analysis questions.