Investors price AI stocks at a premium because those equities hedge a hypothetical AI 'singularity' that could displace consumption; market incompleteness then inflates valuations and creates a case for government transfers when singularity-driven gains outweigh policy costs.

Main Finding

AI-facing public equities can trade at large valuation premia because investors use them to hedge an uninsurable “AI singularity” that would displace their labor/consumption. Market incompleteness—private AI capital (founder stakes, pre-IPO holdings, future firms) is not tradable—creates this hedging demand and therefore a persistent AI stock premium. The same incompleteness that raises AI valuations can also distort AI development choices; government transfers can substitute for missing markets and, in singularity states with explosive growth, even inefficient transfers can become effective.

Key Points

• Economic mechanism
  • A discrete-time model with a representative (marginal) household and a set of AI owners who hold private, non-tradable AI capital.
  • With probability p each period a singularity occurs: with prob. 1−ξ it yields large positive productivity growth (factor 1+η) and displaces the household (their consumption share multiplies by ϕ∈(0,1)); with prob. ξ it causes extinction (Ct+1 = 0).
  • Public assets: AI stocks (dividends = θt Ct) and non-AI stocks (dividends = (1−θt)Ct). AI owners hold restricted equity that the household cannot buy, so public AI stock is only a partial hedge.
• Asset-pricing result
  • Closed-form expressions for price-dividend (P/D) ratios for AI and non-AI stocks are derived (see Equations (5)–(6) in the paper). P/D depends on p, ξ, η, ϕ, ∆θ (the post-singularity increase in AI dividend share), risk aversion γ, β, and baseline growth g.
  • Existence condition: a weighted growth term Aj must be < 1 for finite prices; otherwise the hedging value diverges (prices blow up).
• Quantitative takeaway
  • Calibrations in the paper show that with a 1% per-period singularity probability, AI stock P/D ratios can be roughly double those of non-AI stocks—consistent with large observed valuation gaps.
  • Extinction risk (ξ > 0) attenuates but does not eliminate the premium: states with big upside are also those with higher existential risk, which narrows the spread.
• Real-side distortions and policy
  • Incomplete markets not only affect prices but can make risk-averse households oppose socially efficient AI development because the uninsurable downside looms large.
  • Broader trading of AI equity would reduce the premium, but is constrained by restricted ownership and the fact that much relevant capital is yet to exist.
  • Government transfers can act as a market-completing substitute, but standard transfers are hampered by deadweight costs. However, if a singularity occurs and growth is explosive, even transfers with large inefficiencies can be feasible—the resource abundance overcomes deadweight losses.
• Paper provenance and limitations (noted by the author)
  • The paper’s text, analysis and code were produced using an automated AI pipeline. The model abstracts from endogenous firm/worker entry, assumes a simple dividend and displacement process, and uses an approximation for post-singularity P/D ratios. These modelling simplifications limit some quantitative precision.

Data & Methods

• Theoretical framework
  • Discrete-time infinite-horizon model extending Gârleanu, Kogan, and Panageas (2012) to a singularity environment and incorporating Jones (2024)-style extinction risk.
  • Representative household with CRRA utility (risk aversion γ > 1, discount β), pricing public assets using the household SDF based on its own consumption (incomplete markets).
  • Singularity process: period probability p; conditional outcomes (growth η and displacement ϕ, or extinction with prob. ξ). AI dividend share θt jumps by ∆θ on non-extinction singularities.
  • Two traded assets (AI and non-AI) whose dividends are proportional to aggregate consumption shares.
• Analytical results
  • Derived closed-form price-dividend ratios (Equations (5) and (6)); characterized the finite-price condition Aj < 1.
  • Propositions on how extinction risk affects premiums and on when households may block efficient AI development.
• Quantitative calibration
  • Calibrated parameters (examples discussed in the paper) produce large AI P/D premia for modest singularity probabilities (e.g., p = 1% per period yields ≈2x AI vs non-AI P/D).
  • Sensitivity checks: premium decreases with higher extinction probability and with greater tradability of AI equity (i.e., less market incompleteness).
• Empirical context
  • Paper motivates model with observed high valuations: S&P 500 Shiller P/D and NASDAQ vs S&P 500 valuation trajectories (figures use Shiller dataset and NASDAQ from FRED).
• Methodological caveats
  • The model treats AI owners as a static group (no explicit entry dynamics), simplifies dividend and displacement processes, and relies on an approximation for post-singularity valuation dynamics. The author flags omitted-literature risks and modeling approximations.

Implications for AI Economics

• For asset pricing and investors
  • Interpreting AI stock premia: part of observed high valuations may reflect hedging demand against singularity/displacement risk, not just higher expected cash flows.
  • Portfolios: publicly traded AI equities serve as the primary available hedge when private AI capital is unavailable; increased supply of tradable AI claims would reduce hedging-induced premia.
• For corporate finance and governance
  • Restricted ownership structures (founder stakes, pre-IPO allocations, non-tradable equity) are economically important: they generate market incompleteness that affects both prices and social outcomes. Policies or market innovations that reduce restrictions could change valuations and incentives.
• For policy and regulation
  • Financial instruments and market design can be part of the toolkit for managing AI downside risk—complementary to alignment, safety, and regulatory approaches—but frictions limit purely private-market solutions.
  • Government transfers can function as substitutes for missing hedging markets. Standard welfare costs limit their use ex ante, but in an explosive-growth post-singularity world transfers can be effective even if inefficient; this creates new normative tradeoffs in designing pre- and post-singularity policy.
  • Policymakers should consider the interaction between financial market structure (tradability of AI claims), public redistribution capacity, and incentives for AI development.
• For debates about slowing vs. accelerating AI
  • Some calls to slow AI may reflect privately optimal behavior under incomplete markets (risk-averse agents unable to hedge downside), not purely misaligned social preferences. Market-completing policies could alter the private social calculus and thus the political economy of AI development.
• Directions for research
  • Empirical tests: relate cross-sectional AI-stock premiums to measures of restricted ownership, investor beliefs about singularity probability, or proxies for households’ exposure to displacement.
  • Model extensions: endogenize entry of AI owners and firms, richer incomplete-market structures (e.g., derivative markets, private equity trading), dynamic corporate-finance choices, and heterogeneous agents.
  • Policy design work: optimal transfer schemes conditional on pre- vs post-singularity states, and mechanisms to expand tradability of AI claims without undermining innovation incentives.

Summary note: the paper is primarily theoretical with quantitative calibrations and emphasizes that market incompleteness, not only fundamentals or expectations of cash-flow growth, can drive large AI equity premia and shape the social desirability and policy responses to transformative AI.