What question did this study set out to answer?

The study aims to understand how AI disruption affects value distribution in semiconductor supply chains, particularly examining the differences between Apex and Midstream firms.

February 17, 2026Open Access

AI-Driven Value Redistribution in Semiconductor Supply Chains: Evidence from ChatGPT's Impact on U.S. and Taiwan Markets

Key Points

The study aims to understand how AI disruption affects value distribution in semiconductor supply chains, particularly examining the differences between Apex and Midstream firms.
Analyzed 136 firms in the U.S. and Taiwan from 1986 to 2025
Conducted difference-in-differences analysis using ChatGPT's launch as a natural experiment
Utilized the Technology Capability Amplification framework to explain value concentration
Apex firms captured valuations 12.4 times higher than Midstream manufacturers in the U.S.
AI demand shocks increased Apex firms' Tobin's Q by +65.7%, while Midstream firms declined by -18.8%
Demonstrated a 5.1x valuation premium for fabless models over vertically integrated firms

Abstract

This study examines how artificial intelligence (AI) disruption reshapes value distribution across semiconductor supply chains, challenging traditional "smiling curve" assumptions that predict manufacturing commands premium valuations. Analysing 136 firms across Taiwan and the United States (1986-2025), we document an extreme U-shaped value distribution where ecosystem-controlling Apex firms (IC designers, IP licensors) capture valuations 12. 4 times higher than capital-intensive Midstream manufacturers (foundries, IDMs) in the U. S. market and 6. 2 times higher in Taiwan. Leveraging ChatGPT's November 2022 launch as a natural experiment, difference-in-differences analysis establishes causal evidence that AI demand shocks amplify pre-existing asymmetries: Apex firms gained +65. 7% in Tobin's Q while Midstream declined -18. 8%, an 84. 6 percentage point divergence. We introduce the Technology Capability Amplification (TCA) framework, demonstrating that AI's computational intensity concentrates value at supply chain endpoints, controlling platform ecosystems (NVIDIA's CUDA, ARM's instruction sets) rather than midstream manufacturing scale. The fabless business model commands a 5. 1x valuation premium over vertically integrated IDMs, driven by superior gross margins (+16. 2 percentage points) and capital efficiency (4x lower CapEx intensity). Cross-country validation confirms mechanism consistency: profitability and capital efficiency drive valuations identically across markets (β coefficients statistically equal, p>0. 19), while magnitude differences reflect Taiwan's geopolitical discount and market maturity. Free cash flow yield analysis reveals localised rather than systemic bubbles, with NVIDIA's Q=40. 65 justified by substantial cash generation despite extreme valuation multiples. These findings challenge the 52 billion CHIPS Act allocations favouring capital-intensive manufacturing, suggesting policy misalignment with value capture realities. Our contribution extends platform ecosystem theory to supply chain contexts, establishes causal AI impact through natural experiments, and provides actionable implications for corporate strategy, investment allocation, and industrial policy.

Read Full Paperexternally

Mark Helpful

Bookmark

Relay

View Full Paper