AI Trading Arms Race Compresses Alpha Signals from Years to Months

The half-life of trading signals has collapsed from 5-7 years to 18 months as artificial intelligence adoption reaches 70% across institutional trading desks, according to new research that maps the mathematical mechanics of alpha decay in algorithmic markets.

The compression follows a precise formula: h(φ)=ln2/[θ+δ(φ)], where signal lifespans decrease in convex fashion as AI adoption φ rises. At current penetration levels, the model suggests traditional quantitative strategies face extinction timelines measured in quarters rather than years.

The Three Channels of Alpha Erosion

The research identifies three mutually reinforcing mechanisms driving this compression. Signal crowding occurs as multiple AI systems simultaneously exploit the same market inefficiencies, rapidly arbitraging away excess returns. Performative signal erosion follows as widespread use of identical or similar algorithms changes the underlying market dynamics that originally created the signal. Red Queen competition emerges as firms continuously upgrade their AI capabilities just to maintain existing performance levels.

These channels operate simultaneously rather than sequentially. A momentum signal that historically generated excess returns for 3-4 years might now face crowding within months, performative degradation as price patterns shift in response to widespread exploitation, and competitive pressure from rival algorithms targeting adjacent alpha sources.

Signal Extinction Cascades

Beyond individual signal decay, the research identifies a critical threshold where the collapse of one signal class triggers accelerated competition for remaining opportunities. This "signal extinction cascade" represents a phase transition in market dynamics rather than gradual degradation.

The mathematics suggest that once AI adoption crosses certain penetration levels in specific asset classes or strategies, the system exhibits non-linear behavior. The death of high-frequency mean reversion signals, for instance, could intensify competition for momentum factors, accelerating their decay below normal baseline rates.

Looking at the current environment, this pattern rings familiar. The quantitative meltdown of August 2007 previewed similar dynamics when crowded trades unwound simultaneously across multiple hedge funds. The difference now lies in the speed and scope—AI systems can identify, exploit, and saturate opportunities at machine rather than human timescales.

Conflicting Evidence on Market Efficiency

The AI trading revolution presents contradictory evidence about market functioning. Federal Reserve research analyzing language models including Claude 3.7, Claude 3.5, Llama 3, and Nova Pro found varying levels of rational trading performance across different signal treatments, suggesting AI capabilities remain uneven across strategy types.

Academic studies show AI-powered mutual funds significantly outperform human-managed counterparts through superior stock selection and lower turnover ratios. The outperformance stems from reduced transaction costs, enhanced stock-picking capability, and elimination of behavioral biases that plague human portfolio managers.

Yet other research demonstrates that algorithmic trading can undermine efficiency in capital markets, even as separate analysis confirms that algorithmic trading improves market efficiency independent of regulatory reforms.

The contradiction likely reflects the heterogeneous nature of AI implementation across firms and strategies. Early adopters with sophisticated infrastructure capture alpha while late entrants face diminishing returns in increasingly efficient markets.

Implications for Market Structure

The research findings suggest fundamental changes in how institutional investors should approach strategy development and risk management. Portfolio construction models built on historical signal persistence will overestimate future alpha generation. Diversification benefits between quantitative strategies will decline as correlation increases during signal extinction events.

For systematic trading firms, the mathematics point toward shorter strategy lifecycles and higher research intensity requirements. The traditional model of developing a signal, deploying capital, and harvesting alpha over multi-year periods no longer applies. Instead, firms face continuous algorithm refresh cycles and accelerating arms race dynamics.

The broader question concerns market stability. If signal half-lives continue compressing toward months rather than years, and if extinction cascades become more frequent, the foundation for systematic risk-taking erodes. Alpha becomes increasingly transient, potentially concentrating returns among the fastest-adapting participants while leaving traditional asset managers struggling to justify fees.

The decline in skilled human fund managers, with research showing their proportion "substantially dropping and almost non-existent in the new millennium," compounds this dynamic. As human expertise atrophies, market efficiency increasingly depends on algorithmic competition rather than fundamental analysis.

The Endgame Scenario

The mathematical models suggest an eventual equilibrium where AI adoption approaches 100% and signal half-lives asymptotically approach zero. In this limit case, excess returns exist only in the brief moments between signal discovery and algorithmic exploitation—a market approaching perfect efficiency through machine-mediated arbitrage.

Whether markets can function effectively in such an environment remains an open question. The research provides the mathematical framework for understanding current dynamics but offers limited guidance on optimal policy responses or market structure adaptations.

What seems certain is that the current transition represents more than incremental technological advancement. The compression of signal half-lives from years to months marks a phase change in market behavior, with implications extending far beyond individual trading strategies to the fundamental economics of active management and price discovery.