How to Research Unnamed Major Customers in a 10-K
Published September 5, 2026
Altsets
Research by Altsets Research
Use Nvidia's anonymous 22% and 14% customer disclosures to build an evidence-based candidate process without assigning filing percentages to named graph edges.
Dataset reference:Altsets supply chain relationships dataset covers 115k+ entities, 550k+ relationships, and history since 2006 across the full dataset; individual analyses use a narrower scope.
Key findings
- Nvidia publicly discloses large anonymous direct-customer concentration while named supply chain relationships provide candidate counterparties and channel context, not automatic identities.
- A filing concentration percentage should not be attached to a named relationship unless period, customer role, accounting basis, and additional evidence support the identification.
To research an unnamed major customer in a 10-K, use named relationship data to build candidates, then test each candidate against dated evidence. Do not assign an anonymous revenue percentage to whichever famous customer appears in the graph. A concentration disclosure establishes economic importance; it does not establish identity.
Nvidia's fiscal 2026 annual report illustrates the problem. It reported one direct customer at 22% of total revenue and another at 14% without naming either. The supplied Altsets graph separately maps downstream relationships with Amazon, Microsoft, Super Micro Computer, Quanta Computer and Samsung Electronics. Those names create an investigation set, rather than a decoding key for the filing.
The displayed metrics make the limit clearer. Nvidia to Quanta is estimated at 0.53% of Nvidia revenue and 1.76% of Quanta's cost base. Nvidia to Samsung is estimated at 0.04% of Nvidia revenue and 0.10% of Samsung's costs. Neither supplier-revenue estimate matches the anonymous 22% or 14% disclosure. That mismatch is a reason to preserve separate observations, not search for an arithmetic explanation that forces them together. Differences in period, channel role and measurement basis must be resolved before a named identity can be claimed.
Start an evidence record with the anonymous disclosure intact
Record the filing, economic period, concentration percentage, accounting measure and whether the customer is direct or indirect. Keep the anonymous label exactly as reported, without treating it as a permanent company identifier. The Nvidia fiscal 2026 annual report is the source for the 22% and 14% revenue observations.
A named customer row should initially sit beside that disclosure, not replace it. Store the commercial direction as Nvidia supplying the candidate customer. Supplier revenue percentage belongs to Nvidia's sales base; customer cost percentage belongs to the buyer's cost base. A customer-cost figure cannot identify a revenue-concentration percentage merely because the numeric values look similar.
A direct buyer and an economic end user may be different companies
Nvidia's filing distinguishes direct purchases from purchases through other customers. It identifies several possible direct-customer categories, including distributors, device makers, cloud providers and system integrators. Consequently, an end user's economic demand need not identify the legal buyer behind a large direct-customer revenue disclosure.
This distinction changes the candidate test. A well-known cloud company can be relevant to ultimate demand while an intermediary handles the purchase. Conversely, a named direct relationship need not account for the full demand associated with a downstream brand. The graph does not establish those channel allocations without supporting evidence.
The direct and indirect customer guide addresses aggregation once channel structure is supported. The present task is narrower: deciding what can be said about an anonymous customer's identity.
What would justify moving a candidate toward identification?
Look for customer-specific evidence that connects the named entity to the relevant sales channel and period. Official statements, purchase commitments, product disclosures or contract descriptions may strengthen a match. Receivables concentration can add context, but it measures a different financial balance and should remain separate from recognized revenue.
Product compatibility, geography and shipment or procurement evidence can narrow the set. None should be promoted to proof simply because it agrees with a plausible story. A supplier's general product portfolio does not establish who bought those products in the disclosed concentration amount. A similar percentage from a different period or an estimated dataset does not establish identity either.
The decision record should state the strongest supported conclusion. "Named commercial relationship" means the graph connects the companies. "Candidate for the anonymous customer" adds a hypothesis requiring further evidence. "Identified customer" needs direct support for the mapping. If the evidence stops at the candidate stage, keep it there.
Use mismatches to choose the next source
The small named Quanta and Samsung percentages do not explain Nvidia's anonymous major-customer disclosure. They do show where the supplied map provides quantified named relationships and where it does not. That changes the research queue: investigate unmeasured named candidates and their channel roles rather than relabeling measured small edges as the major buyer.
A complete candidate assessment compares period, entity, directness and accounting basis. Keep unresolved dimensions visible. The named Nvidia customer analysis addresses the limited concentration conclusion supported by the two quantified edges; it is not an identity claim about the annual report.
Track concentration across filings without inventing customer continuity
An anonymous label is local to its filing unless evidence establishes continuity. Nvidia's fiscal 2025 disclosures show three direct customers at 12%, 11%, and 11% of annual revenue. Its third quarter of fiscal 2026 shows four direct customers at 22%, 15%, 13%, and 11% of quarterly revenue. The disclosures warn that customer references in one period may represent different customers in another. A chart connecting successive values for "Customer A" can therefore connect different companies.
Preserve each observation's filing date, economic period, annual or quarterly denominator, direct or indirect channel, accounting measure, and anonymous label. Record a proposed identity or continuity mapping separately, together with the dated evidence supporting it. Moving above or below a disclosure threshold does not itself establish entry or exit, and a quarterly share should not be stitched into an annual series as though the denominators matched.
Identity uncertainty does not prevent concentration analysis. The number of customers above a threshold, largest disclosed share, and changes in channel or receivables concentration remain observable without naming every buyer. Keep those conclusions distinct from a claim that a particular customer's purchases increased. A similar percentage in a named relationship estimate remains a clue requiring reconciliation of date, denominator, channel role, and measurement basis.
Historical evidence must also have been available at the historical research date. Today's relationship graph cannot identify an anonymous buyer in an older filing without contemporaneous support. Preserve earlier unresolved hypotheses when new evidence arrives, instead of rewriting the historical record as though the name had always been known. The point-in-time backtesting guide explains that availability constraint.
A later filing is new evidence, not a retrospective answer
Revisit candidate identities as new customer-specific disclosures appear, but preserve what was supported at each research date. Establishing continuity across periods requires additional evidence beyond identifying candidates in one filing.
The defensible output here is a dated evidence trail, with anonymous disclosures, named relationships and unresolved identity hypotheses in separate fields. Altsets methodology explains the estimates and point-in-time constraints that govern that trail.
Sources
Related research
Methodology
Cite this research
Altsets Research. "How to Research Unnamed Major Customers in a 10-K." Published September 5, 2026. https://www.altsets.com/research/unnamed-major-customers-supply-chain-data