Investment alpha is dying faster than ever, and artificial intelligence is the prime suspect. According to research by Shuchen Meng and Xiupeng Chen of New York University, the mass adoption of AI strategies is becoming an act of collective suicide in the markets. The scholars have mathematically proven that as funds deploy identical models, they endogenously destroy the very inefficiencies they intended to exploit. This is no mere academic speculation: the researchers derived an "Alpha Half-Life Theorem," confirming that the lifespan of a profitable signal shrinks non-linearly as algorithmic density increases.
The Three Channels of Signal Degradation
Returns are collapsing due to the "three horsemen" of degradation: signal crowding, performative erosion, and the "Red Queen" race. Crowding stems from an algorithmic monoculture: when hundreds of systems train on the same data—SEC filings and price charts—their trading patterns inevitably converge. Performative erosion, as defined by Meng and Chen, creates a reflexive loop: the very act of executing a trade based on a signal accelerates the arbitrage of that pattern until it vanishes entirely. Finally, the Red Queen race forces funds to spend millions on AI just to stay in place. This is the paradox of intelligent markets: technology designed to provide an edge turns the hunt for excess returns into a zero-sum game. An analysis of 99.5 million records from 13F filings shows that institutional portfolio convergence grew by 42% between 2013 and 2024.
In a state of monocultural equilibrium, pure alpha is identically zero despite massive investments in AI.
The most painful finding of the study is the radical compression of strategy lifecycles. In the pre-algorithmic era, a viable trading signal could sustain an investor for 5 to 7 years. Today, with current AI adoption levels and signal correlations hovering around 0.6, that window has collapsed to 18 months. The formula developed by Meng and Chen, which accounts for natural mean reversion and AI-accelerated decay, clearly demonstrates that every new market participant wielding a neural network shortens the life of signals for everyone else at a progressive rate.
The 18-Month Deadline
The model suggests that the signal half-life is now 18 months, compared to 5–7 years in the pre-algorithmic era.
This acceleration triggers a cascade reaction: once a critical threshold of AI adoption is crossed, the disappearance of one class of signals intensifies the scramble for remaining patterns, destabilizing the entire system. The researchers point to a dangerous trade-off between efficiency and fragility. The level of algorithmic participation required for rapid price discovery far exceeds the level at which the system remains resilient. Markets are effectively becoming "glassy," a theory supported by simulations of the 2010 flash crash.
- AI usage has slashed the average lifespan of a trading signal from 5–7 years to a mere 18 months.
- Institutional portfolio convergence has jumped 42% over the last decade, reducing return variance.
- Market efficiency now comes at the cost of systemic fragility: the risk of flash crashes rises as signals synchronize.
For fund managers and CTOs, this reads like a death sentence for off-the-shelf solutions. As Meng and Chen demonstrate, AI has become a commodity that devalues returns through algorithmic monoculture. To extract alpha in this environment, you must either find data sources far beyond standard datasets or build architectures that avoid the "0.6 correlation zone." Otherwise, your AI investment is not a growth driver, but a costly tax on survival. The era of multi-year strategies is officially over; you are in a sprint where the finish line moves closer every time a competitor hits the "train" button.