Taiwan's June ADP exports declined M/M, but the underlying trend in NVIDIA's AI demand remains robust

Overview

This report evaluates the implications of Taiwan's June export data, released by the Ministry of Finance, for NVIDIA's data center revenue and AI demand trends. The core conclusion is that the M/M decline in Taiwan's ADP exports excluding notebook exports primarily reflects normalization after prior volatility and should not be simply interpreted as deterioration in AI demand; absolute export levels remain high, continuing to support the view that demand for NVIDIA's data center business is strong.

Core views

UBS's core view is positive: June Taiwan ADP exports excluding notebook exports were US$25.2 billion, down 4.0% from US$26.3 billion in May. However, May itself had rebounded 11.5% after a sharp April decline, making the June pullback look more like a short-term correction in pace. Assuming July normally accounts for approximately 37% of F2Q exports, the data imply approximately US$81.9 billion for F2Q, up 16% Q/Q. Although this is below UBS's forecast for NVIDIA's data center revenue, it is consistent with the historical pattern of deviations between Taiwan ADP data and NVIDIA data center revenue.

Analysis framework

The report analyzes the correlation between Taiwan's ADP exports excluding notebook exports and NVIDIA's data center revenue, incorporating monthly and quarterly seasonality, monthly contribution ratios, and historical forecast errors to assess the explanatory power of export data for NVIDIA's performance. The analysis does not equate Taiwan ADP exports with NVIDIA revenue, but explicitly discusses differences in scope: ADP primarily covers servers, excludes products in which chips are exported first and assembled in other regions, excludes networking equipment, and also includes traditional servers and competitors' products.

Methodology notes

• Operating data trackingCorrelation analysis between Taiwan ADP exports and NVIDIA data center revenue

  Use Taiwan ADP exports excluding notebook exports as an external tracking indicator of AI server demand.

  This indicator has some predictive value for NVIDIA data center revenue, but cannot be mapped one-to-one to NVIDIA revenue due to factors including product scope, assembly locations, exclusion of networking equipment, and the inclusion of competitors' products.

• Valuation methodsP/E multiple valuation

  Derive the target price based on an 18x P/E multiple and C2027E EPS of US$15.37.

  UBS applies an 18x valuation multiple to expected calendar-year 2027 EPS of US$15.37 to derive a US$280 target price for NVIDIA.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• NVIDIA CORP (NVDA.O)

  Core covered security; Taiwan ADP export data are used as an external validation indicator for its data center and AI demand.

  Strengths

  AI-driven demand remains healthy, absolute export levels are high, and UBS maintains a Buy rating with a target price materially above the current price.

  Weaknesses

  The scope of Taiwan ADP exports does not fully cover NVIDIA data center revenue, and monthly data are highly volatile.

  Comparison

  Compared with NVIDIA data center revenue, Taiwan ADP exports have less coverage of chips assembled in other regions and networking equipment, while including traditional servers and competitors' products; therefore, they are related but not equivalent.

  Risks

  Risks from competition, the semiconductor cycle, differences in data scope, and the market misinterpreting single-month export normalization.

Key data

• June Taiwan ADP exports excluding notebook exportsUS$25.2BDown 4.0% M/M from US$26.3B in May.
• Normal seasonality-2.8% M/MJune's actual M/M performance was weaker than normal seasonality.
• May M/M change+11.5% M/MSignificantly above normal seasonality of +1.7%, partially recovering from the sharp April decline.
• April M/M change-28.4% M/MThe preceding sharp decline provided the backdrop for the May rebound and June normalization.
• Implied F2Q scaleApproximately US$81.9B / +16% Q/QBased on the assumption that July normally accounts for 37% of total F2Q exports.
• Target priceUS$280.00Based on an 18x P/E multiple and C2027E EPS of US$15.37.
• Current priceUS$202.78Price as of July 9, 2026.
• Forecast price appreciation38.1%Forecast price appreciation disclosed in the report table.
• Forecast total return38.6%Includes a forecast dividend yield of 0.5%.

Impact & implications

The implication for investment judgment is that weakness in a single month's Taiwan export data is insufficient to overturn the core thesis of strong NVIDIA AI data center demand. If absolute levels remain high in subsequent months, they will continue to support assumptions of NVIDIA revenue growth and elevated valuation; however, given the structural differences in scope between Taiwan ADP and NVIDIA data center revenue, the market should not extrapolate company performance directly from a single month's M/M export change.

Risks

• Taiwan ADP exports and NVIDIA data center revenue differ significantly in scope, and substantial forecast errors have sometimes occurred historically.
• GPU and professional visualization products face pressure from competitors such as AMD.
• ARM application processors and professional GPU computing may face competition from alternative technology approaches.
• Semiconductor industry revenue trends are linked to corporate earnings and the macroeconomic cycle; growth could come under pressure if the economic recovery is weaker than expected.
• Volatility in monthly export data could lead the market to misjudge the pace of AI demand.

What to watch

• Whether July Taiwan ADP exports excluding notebook exports are consistent with an approximately 37% contribution to F2Q.
• Whether absolute Taiwan ADP export levels remain near historical highs in subsequent months.
• The deviation between NVIDIA's data center revenue in the next quarter and the scale implied by Taiwan ADP data.
• Changes in AI server demand, networking equipment, and the supply chain for chips assembled in other regions.
• Competitors' progress in GPU, professional computing, and AI acceleration.