High AI reliance among listed new-energy and auto firms correlates with greater financial risk rather than improved financial health; firms that effectively convert R&D into patents or intelligent equipment avoid the worst effects.

Main Finding

Using a 2013–2023 panel of A‑share listed new energy vehicle and automobile manufacturers, the paper finds that higher firm-level AI dependency (as disclosed in corporate reports) is associated with weaker financial safety — i.e., greater corporate financial risk. This adverse effect is attenuated where firms have stronger knowledge/intelligent‑equipment output and higher patent transformation (better R&D→commercialization efficiency). Investor confidence is identified as a mediating channel, and innovation efficiency moderates the AI→financial risk relationship.

Key Points

• Primary result: Higher AI dependency does not improve—and on average worsens—financial health for NEV/auto manufacturers in the sample period.
• Mechanisms:
  • Investor confidence mediates the relationship: AI reliance can reduce investor trust and thereby raise financing costs/liquidity risk.
  • Innovation efficiency moderates effects: firms that convert R&D into patents, intelligent equipment output, or other tangible innovation outcomes suffer less (or avoid) the negative financial consequences of AI dependence.
• Interpretation: AI investments are capital‑intensive, have uncertain short‑term returns, and can crowd out more productive R&D or operational spending, producing a kind of “intelligent trap” for firms lacking efficient R&D commercialization.
• Robustness: The paper reports moderation and heterogeneity tests supporting the role of knowledge output and patent transformation in mitigating risks.
• Policy/business implication emphasized by authors: NEV firms should prioritize R&D efficiency and ensure innovation leads to stable returns to avoid AI‑driven financial distress.

Data & Methods

• Sample: Listed new energy vehicle and automotive manufacturing companies on Shanghai and Shenzhen A‑share markets, 2013–2023.
• Data sources: Corporate financial statements, East Money Information sector lists, GUOTAIAN (CSMAR) database, and Wind platform.
• Key variables (paper description):
  • AI dependency: measured from corporate disclosures/financial reports (text/disclosure‑based indicator as presented in the firms’ periodic reports).
  • Financial risk / financial safety: composite/standard financial indicators (liquidity, leverage, cash‑flow metrics) used to assess corporate financial health.
  • Mediator: investor confidence (proxied by market/investor signals derived from disclosures/market data).
  • Moderator(s): indicators of innovation efficiency — patent transformation, knowledge/intelligent equipment output, R&D conversion measures.
• Empirical strategy:
  • Panel regression models linking AI dependency to financial safety.
  • Mediation analysis to test investor confidence as an intermediate channel.
  • Moderation and heterogeneity tests to examine how innovation efficiency and knowledge/intelligent‑equipment outputs alter the relationship.
  • Sample handling: observations with large data gaps excluded; minor missing values imputed by mean; no winsorization applied (authors argue for preserving variation across firms).
• Limitations noted or implied: disclosure‑based AI measures, potential endogeneity between AI investment and financial outcomes, and measurement heterogeneity in corporate AI adoption.

Implications for AI Economics

• Measurement caution: Firm‑level AI exposure estimated from disclosures can be informative but may overstate capabilities; researchers should triangulate disclosure text measures with tangible proxies (patent counts, intelligent equipment output, robot installations) and consider disclosure bias.
• Heterogeneous returns to AI: The paper highlights important heterogeneity — AI investment is not uniformly beneficial. Returns depend critically on firms’ ability to commercialize R&D and integrate AI into productive processes.
• Financial channel awareness: AI investments can raise short‑term financial fragility through higher upfront costs and reduced investor confidence. Modeling AI adoption effects needs to account for financing frictions and market sentiment.
• Policy and managerial design: To avoid an “intelligent trap,” firms and policymakers should support mechanisms that improve R&D conversion (patent commercialization, standards, complementary physical capital), and provide financing instruments that accommodate longer innovation horizons.
• Future research directions:
  • Causal identification: exploit exogenous variation in AI exposure (policy shocks, regional AI infrastructure rollouts, supply shocks) to disentangle causality.
  • Granular measures: combine text analysis with machine‑readable measures (capex on intelligent equipment, hires with AI skillsets, patent citation quality).
  • Cross‑sector comparison: test whether the negative short‑term financial effects generalize beyond capital‑intensive sectors like auto/NEV.
  • Firm heterogeneity: further explore roles of ownership (SOE vs private), firm size, and pre‑existing technological capabilities in moderating AI’s financial impacts.

Summary takeaway: AI adoption in capital‑intensive, rapidly evolving sectors can raise financial risk when R&D and AI investments do not efficiently translate into commercialized technology and stable cash flows; increasing R&D conversion efficiency and signaling credible technological outcomes are key to avoiding an “intelligent trap.”