Customer Geography Mix & China Revenue Exposure (Nvidia)
I have comprehensive geographic revenue data from Nvidia's FY24, FY25, and FY26 10-K filings. Let me build a chart-ready dataset and present the full picture.
I need to use df_668 (the loaded dataframe). Let me skip the compute and go directly to the chart and presentation.
Nvidia — Customer Geography Mix (FY22–FY26, $B)
Revenue is reported by customer billing location (FY22–FY25) and was reclassified in Q3 FY26 to customer headquarters location. Prior-year figures were recast.
| Region | FY22 | FY23 | FY24 | FY25 | FY26 |
|---|---|---|---|---|---|
| United States | 4.3 | 8.3 | 27.0 | 61.3 | 149.6 |
| Taiwan | 8.5 | 7.0 | 13.4 | 20.6 | 42.3 |
| China (incl. Hong Kong) | 7.1 | 5.8 | 10.3 | 17.1 | 19.7 |
| Singapore | n/d | 2.3 | 6.8 | 23.7 | n/d |
| Other | 6.9 | 3.6 | 3.4 | 7.9 | 4.3 |
| Total | 26.9 | 27.0 | 60.9 | 130.5 | 215.9 |
Key observations
- US share has surged from 16% (FY22) to ~69% (FY26), driven by explosive demand from U.S.-headquartered hyperscalers and AI model makers for Compute & Networking products.
- Taiwan (mostly OEM/ODM hubs for Foxconn, Wistron, Quanta serving US hyperscalers) is the second-largest geography. In FY26, Nvidia disclosed that ~76% of Taiwan-headquartered Data Center revenue was actually shipped to end-customers in the US and Europe — i.e., Taiwan revenue is largely a pass-through for US demand.
- Singapore (revenue 18% in FY25) is primarily an invoicing hub — Nvidia noted physical shipments into Singapore were <2% of total revenue. Singapore was rolled into "Other" or recast out in the FY26 recast.
- China revenue peaked at ~$25B in FY25 (recast FY26 headquarters basis) before collapsing in FY26 to $19.7B despite strong fiscal growth, as export controls took effect.
China revenue exposure — a sharp step-down
| FY23 | FY24 | FY25 | FY26 |
|---|---|---|---|
| 21% of DC rev / 21% total | 14% of DC rev / 17% total | mid-teens % DC | ~9% total |
- USG export controls (H100/H800/A100 → then expanded to H20/L40S/RTX 4090/Blackwell) require licenses for China shipments. To date Nvidia has received no meaningful licenses for restricted products.
- Nvidia shipped degraded China-specific SKUs (H20), but in Q1 FY26 took a $4.5B H20 inventory & purchase obligation charge after April 2025 restrictions.
- A small August 2025 license yielded ~$60M of H20 revenue; a Feb 2026 H200 license has generated no revenue to date and imposes a 25% US-import tariff.
- China antitrust regulators (Sep 2025) preliminarily concluded Nvidia's export-control compliance discriminated against Chinese customers, violating Mellanox-merger commitments — exposure to penalties or operational restrictions.
- As of end-FY26, Nvidia states it is effectively foreclosed from China's data center compute market and "unable to create a competitive China product acceptable to both USG and PRC."
- China is now a mid-to-high single-digit % of Data Center revenue — down from ~19% at the start of the AI cycle (FY23).
Bottom line: Nvidia's revenue base is now decisively US-anchored (~69% of FY26 revenue), Taiwan functions as a contract-manufacturing funnel for US demand, and direct China exposure has been structurally compressed by export controls, with policy risk elevated on both sides of the Pacific.