More Than a Quarter of Nvidia's Revenue Now Comes From Customers Based in Taiwan. Should Investors Be Worried?

Nvidia billed $27.0B, or 28% of fiscal Q2 revenue, to Taiwan-headquartered direct customers, up from $8.9B a year earlier and roughly $12B in the prior quarter. Total revenue reached $96.2B, up 106% year over year, with data center contributing about 92%; management forecast $108.0B for fiscal Q3. The filing also highlights concentration risk: one direct customer represented 16% of quarterly revenue, while Taiwan is central to both Nvidia's customer-routing and manufacturing supply chain.

Analysis

The relevant signal is not regional end-demand but a rising dependence on a small number of AI-server integrators as the product mix shifts toward rack-scale systems. That raises quarterly revenue-recognition, acceptance-testing and inventory-timing risk: a delayed cluster deployment can move billions of revenue across quarters without changing underlying hyperscaler capex. NVDA's valuation will therefore be increasingly sensitive to backlog conversion and receivables quality, not merely aggregate AI spending.

TSM gains from sustained accelerator volume, but the greater second-order beneficiary may be the Taiwan server-manufacturing chain, where system integration content rises faster than chip content per deployment. Conversely, NVDA bears more execution risk as it moves closer to full-system delivery: networking availability, power/cooling constraints, and customer-site commissioning can become bottlenecks that limit reported growth even if GPU demand remains intact. This should widen the gap between AI demand headlines and near-term shipment outcomes over the next 1-3 quarters.

The market may be underpricing the bilateral Taiwan exposure as a correlation risk rather than a discrete geopolitical tail. A disruption would simultaneously constrain NVDA supply, impair the intermediary channel through which a growing share of product is invoiced, and delay customer deployments; diversification of wafer capacity alone would not solve server assembly and test dependencies. The nearer-term falsifier is simpler: if next-quarter revenue converts cleanly while receivables, inventory and gross margin remain controlled, concentration is operationally manageable rather than a thesis break.

Near term, treat any pullback driven solely by the geographic disclosure as noise unless it coincides with weaker forward supply commitments or a guide-down. Over 6-18 months, the key debate is whether rack-scale complexity reduces NVDA's incremental gross-margin resilience and gives vertically integrated hyperscalers more leverage in pricing and architecture choices.

Key Decisions for Investors

• Maintain NVDA exposure but avoid adding ahead of the next earnings print; add only if management reiterates demand visibility and receivables/inventory growth remains below revenue growth. Thesis risk: a guide-down or material gross-margin deterioration tied to system integration would warrant reducing exposure.
• Pair trade for a 3-6 month horizon: long TSM / short NVDA in equal volatility-adjusted dollars. This isolates fabrication-volume durability from NVDA's customer-concentration and rack-delivery execution risk; close if NVDA demonstrates sustained margin expansion alongside clean backlog conversion.
• Establish a watchlist rather than a position in Taiwan AI-server assemblers and component suppliers; require independently verifiable order-backlog, capacity-utilization and margin data before buying. The likely upside is in integration content, but direct-customer billing data alone does not identify the economic beneficiary.
• Use NVDA downside protection around the next two earnings cycles via 3-6 month put spreads rather than outright short exposure. The asymmetric catalyst is a deployment-delay narrative that compresses the revenue multiple despite intact long-term AI capex; invalidate the hedge if guidance and system-delivery metrics accelerate together.