How to Avoid Double Counting Direct and Indirect Customers
September 6, 2026
Altsets
Research by Altsets Research
Classify legal buyers, channel partners, and end customers before adding customer exposure so the same underlying demand is not counted more than once.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- Nvidia explicitly distinguishes direct customers such as ODMs, OEMs, distributors, cloud providers, and system integrators from indirect customers that can purchase through those direct customers.
- A flat customer list can therefore double count one economic demand stream unless channel roles and multi-step paths are classified before exposures are aggregated.
Avoid double counting by keeping direct buyers and indirect end customers in separate layers and never adding their exposures unless the underlying demand is demonstrably distinct. A supply-chain graph can correctly contain both the company that buys a product and the company whose end demand causes the purchase; treating them as independent can count the same demand twice. Nvidia's customer disclosures make this problem unusually clear.
The company distinguishes direct customers from indirect customers. Direct customers can include ODMs, OEMs, distributors, cloud providers, AI model makers, and system integrators. Indirect customers primarily buy through those direct customers. That means one end-market demand stream can appear at several nodes in the network.
A direct customer is not always the final demand source
An ODM can buy Nvidia products directly and build systems for a cloud provider. The cloud provider can therefore matter economically even when the ODM is the legal purchasing counterparty. Nvidia explicitly says some indirect customers purchase through system integrators and distributors.
The company also estimates indirect-customer revenue using information such as purchase orders, product specifications, internal sales data, and other sources. This is a channel-mapping problem, not just a customer-list problem.
The supplied Nvidia graph contains several channel roles
The supplied Altsets network shows downstream relationships involving Quanta Computer, Amazon, Microsoft, Super Micro Computer, and Samsung Electronics. Quanta and Super Micro Computer operate in server and system manufacturing. Amazon and Microsoft operate large cloud businesses.
Samsung spans several technology and manufacturing businesses. The graph establishes named commercial relationships. It does not, by itself, tell us whether one relationship sits upstream of another inside the same final purchase. That is exactly why summing all visible customer exposures can be dangerous.
Double counting can create a fake concentration number
Imagine a simplified chain: Nvidia sells to a system manufacturer. That system manufacturer builds infrastructure for a cloud provider.
If an investor separately estimates exposure to the manufacturer and the cloud provider, then adds both percentages as if they represent independent demand, the same servers can enter the calculation twice. The graph can be correct at both edges. The aggregation can still be wrong.
Direct and indirect customer disclosures should remain separate
Nvidia's fiscal 2026 filing said one direct customer represented 22% of total revenue and another represented 14%. The company separately said certain indirect customers can individually represent 10% or more of revenue. Those disclosures are not designed to be added into one concentration table.
The direct-customer percentages refer to legal purchasing relationships. The indirect estimates attempt to trace demand farther through the channel. An investor should preserve that distinction.
Channel role matters more than company fame
A well-known cloud company can be economically important without being the direct buyer on a particular transaction. A less-famous ODM can be the direct customer while serving that cloud company. This is why a brand-name customer list can mislead. The important question is: Where does this company sit in the transaction chain?
Possible roles include:
- component supplier;
- ODM;
- OEM;
- distributor;
- system integrator;
- cloud provider;
- end enterprise;
- public-sector end customer.
The same company can even occupy different roles in different transactions.
Relationship percentages need channel context
Suppose a named direct relationship carries a supplier-revenue percentage. That metric measures the mapped commercial relationship. It should not automatically be combined with an estimated indirect end-customer percentage from a filing.
The data may have different denominators, periods, estimation methods, legal counterparties, and levels of the channel. The safest aggregation rule is to sum only relationships that represent mutually exclusive economic demand.
A graph can help prevent the mistake
Instead of treating every customer as a flat list, model the channel as a path. For example: supplier to system builder to cloud provider to enterprise demand.
The exact path needs evidence. But once roles are classified, the investor can decide which node belongs in a revenue-concentration calculation and which nodes belong in end-demand analysis. That preserves both pieces of information without adding them together blindly.
When double counting matters most
The risk is especially high in industries with contract manufacturing, distribution, systems integration, cloud reselling, OEM channels, automotive tiering, electronics assembly, and wholesalers. The more layered the distribution chain, the less useful a flat customer list becomes.
A repeatable no-double-count workflow
- Map the named customer relationships.
- Classify each customer's channel role.
- Identify which customers are legal direct buyers.
- Identify which customers are estimated end-demand sources.
- Trace known multi-step paths.
- Do not add direct and indirect exposure unless the demand sets are mutually exclusive.
- Keep legal-counterparty concentration separate from economic end-demand concentration.
- State when a path is inferred rather than confirmed.
- Reconcile channel roles with the customer's own filings.
The customer-headquarters versus end-demand guide explains how channel structure can distort geographic exposure. The unnamed-major-customer guide explains how named relationships can help investigate anonymous filing disclosures without forcing an identity match. For relationship methodology, read the Altsets supply-chain data methodology.