Why Customer Headquarters Can Mislead Supply-Chain Exposure
August 24, 2026
Altsets
Research by Altsets Research
Separate direct-customer headquarters from production geography and end-demand geography instead of treating one country field as the location of the economic exposure.
Data used:Altsets Supply Chain Intelligence: 90k+ entities, 400k+ relationships, 20+ years of history.
Key findings
- Nvidia disclosed that 76% of fiscal 2026 Data Center revenue from Taiwan-headquartered customers was attributed to end customers based in the United States and Europe.
- Customer headquarters, manufacturing location, shipping origin, and end-demand geography are separate layers and should be matched to the event being analyzed.
Customer headquarters can be a poor proxy for where the underlying demand actually comes from. Nvidia's own filings provide a striking example. For fiscal 2026, Nvidia said 76% of Data Center revenue from Taiwan-headquartered customers was attributed to end customers based in the United States and Europe.
That means a geographic revenue table can make Taiwan look like the demand center even when much of the economic end demand is elsewhere. Supply-chain relationship data helps separate those layers.
Billing or headquarters geography answers a legal-counterparty question
A direct customer may be headquartered in Taiwan because it is an ODM, OEM, or manufacturing partner. Revenue can therefore be recorded against the direct counterparty's geography. That does not mean the final server, cloud service, or AI workload is consumed in Taiwan.
Nvidia changed its geographic presentation to use customer headquarters because billing locations could be even less representative. The company still cautions that direct-customer geography and end-customer geography can differ materially.
The relationship graph shows why
The supplied Nvidia network contains relationships with both: Taiwan-based system or manufacturing companies such as Quanta Computer and US cloud and technology companies such as Amazon and Microsoft. Nvidia's own ecosystem materials separately describe cloud providers such as Amazon Web Services and Microsoft Azure, while also listing system manufacturers including Quanta Cloud Technology and Super Micro Computer. Those are different roles in the same infrastructure stack. A geographic analysis that stops at the first legal counterparty can therefore miss where the downstream demand is actually generated.
This is not the same problem as factory location
There are at least three geographic layers in supply-chain research:
- company headquarters
- production or shipping location
- end-customer demand location
They answer different questions. A Taiwan-headquartered ODM can assemble systems for US or European end customers. A US-headquartered cloud provider can operate data centers globally.
A supplier can manufacture components in yet another country. Collapsing these layers into one "country exposure" number creates false precision.
Why this matters for geopolitical analysis
Suppose an investor is measuring exposure to a regional demand slowdown. Customer-headquarters geography may overstate the connection if the customer primarily serves end users elsewhere. Now suppose the investor is measuring manufacturing disruption.
End-customer geography may be irrelevant if the production facility sits in the affected region. The right geography depends on the event. Supply-chain data is most useful when each relationship is assigned the geographic layer relevant to the question.
Why this matters for tariff analysis
Tariffs add another layer: country of origin. A customer's headquarters does not determine the customs origin of the product. An ODM headquartered in Taiwan can manufacture in several countries.
A US customer can buy goods produced elsewhere. That is why a tariff model needs product origin and classification rather than a customer-address field. The tariff stress-test guide covers that policy layer.
A better geography model
A robust relationship record can attach multiple geographic attributes:
| Geography field | What it helps answer |
|---|---|
| Supplier headquarters | Corporate and jurisdictional exposure |
| Customer headquarters | Direct counterparty geography |
| Production facility | Manufacturing disruption |
| Shipping origin | Logistics and customs |
| End customer | Demand location |
| End market | Economic demand driver |
Not every relationship will have every field. The important thing is to avoid silently substituting one for another.
Nvidia is a particularly useful case
Nvidia's fiscal 2026 disclosure makes the problem measurable. If 76% of Data Center revenue tied to Taiwan-headquartered customers was attributed to US and European end customers, then direct-customer geography and economic-demand geography were materially different. That is exactly the type of distinction a relationship network can help investigate.
The graph provides named counterparties. The filing provides the channel warning. Together they produce a better geographic interpretation than either source alone.
A repeatable geographic-demand workflow
- Choose the event you are trying to model.
- Identify direct customer relationships.
- Record customer headquarters.
- Identify ODM, OEM, distributor, cloud, and system-integrator roles.
- Resolve production geography where relevant.
- Resolve end-customer geography where possible.
- Keep the layers separate.
- Weight relationships only after the geography matches the research question.
- State which geography the final exposure number represents.
The geographic supplier-exposure guide deals with supplier-country screening. This article covers the opposite side of the problem: why customer headquarters can differ from the location of the demand ultimately driving the relationship.
For relationship and historical-data methodology, read the Altsets supply-chain data methodology. Browse Supply-Chain Data Use Cases for other geographic and event-driven workflows.