AI Is Making Stock Trading Signals Expire Faster—Here's Why

Trading signals used to last 5 to 7 years. Now they last about 18 months. The reason: artificial intelligence adoption has reached 70% across institutional trading desks, according to new research analyzing how algorithmic markets compress profitable trading opportunities.

To understand what's happening, think of a "trading signal" as a repeatable pattern—say, a tendency for tech stocks to rise on certain days. A trader spots it, builds a strategy around it, and profits for years. But when AI spreads across the industry, that same pattern gets spotted by dozens of machines simultaneously. The opportunity closes fast.

The underlying mathematics follow a simple rule: as more AI systems adopt a signal, the time it remains profitable shrinks in an accelerating curve. At today's penetration levels, the model suggests that traditional quantitative strategies—the math-based approaches that once thrived—face expiration timelines measured in months rather than years.

Three Ways AI Trading Destroys Profitable Opportunities

The research identifies three channels working together to compress these opportunities.

Signal crowding happens when multiple AI systems spot the same inefficiency and exploit it at the same time, arbitraging the profit away almost instantly. Performative signal erosion follows because the widespread use of identical algorithms actually changes the underlying market patterns that created the signal in the first place—the market adapts to the machines hunting it. Red Queen competition is the relentless arms race: firms continuously upgrade their AI just to maintain the performance they already had, let alone improve it.

These three channels operate in parallel, not one after the other. A momentum signal—a tendency for rising stocks to keep rising—might once have been profitable for three or four years. Now it faces crowding within months, then loses effectiveness as prices shift in response to widespread algorithmic trading, then gets squeezed harder as competitors pile in with their own upgraded algorithms.

Cascading Collapses

The research points to something sharper than slow decay: a threshold effect. Once AI adoption hits certain levels in a given market or strategy type, the system shifts behavior abruptly rather than gradually. This "signal extinction cascade" is a phase transition—the difference between water gradually cooling and ice suddenly forming.

The mathematics suggest that when one signal class dies off, firms rush to compete for what remains, accelerating the decay of the next opportunity. The death of high-frequency mean reversion signals (the practice of betting that prices will snap back to average levels) could push competition into momentum factors, for example, speeding their decline beyond the normal baseline.

This pattern echoes the quantitative meltdown of August 2007, when crowded hedge fund trades unwound simultaneously. But there's a crucial difference now: AI systems identify, exploit, and saturate opportunities at machine speed rather than human speed. That means the whole cycle compresses.

The Paradox: Is AI Making Markets More Efficient—or Less?

The evidence on this point is genuinely contradictory, and worth taking seriously.

Federal Reserve research tested how well different AI language models performed at trading based on various signals. The results varied considerably, suggesting that AI's ability to make money differs widely depending on the type of strategy—some work better than others.

Separately, academic studies found that AI-powered mutual funds significantly outperform human-managed funds through better stock selection and lower trading costs, suggesting machines are genuinely better at finding value.

Yet other papers claim that algorithmic trading can undermine market efficiency, even as separate research insists that algorithmic trading improves efficiency.

The real story is likely this: early adopters with sophisticated AI infrastructure capture profits while latecomers face diminishing returns in increasingly efficient markets. Not everyone benefits equally. The sophistication and resources you can deploy matter enormously.

What This Means for Portfolio Managers

The research findings shift the ground for how professional investors should think about building portfolios.

Portfolio construction models built on the assumption that profitable patterns persist for years will overestimate the alpha—the excess return—they can actually generate. Diversification benefits between different quantitative strategies will shrink as those strategies become correlated during signal extinction cascades.

For systematic trading firms—the ones that rely on algorithmic strategies—the math points toward a grinding treadmill. Strategy lifecycles get shorter. The research intensity required climbs. The old model—develop a profitable signal, deploy money, harvest returns for years—no longer works. Instead, firms enter continuous algorithm refresh cycles and accelerating competition just to stay even.

There's a broader concern worth flagging. If signal half-lives keep compressing toward months, and if extinction cascades become more frequent, the economic foundation for active trading erodes. Returns get concentrated among the fastest-moving competitors. Traditional asset managers struggle to justify their fees. And research shows that the number of skilled human fund managers has been "substantially dropping and almost non-existent in the new millennium"—so as human expertise fades, market functioning increasingly depends on algorithmic competition rather than fundamental analysis.

The Long-Term Endpoint

The mathematics point toward an eventual state where AI adoption approaches near-100% and profitable signals disappear almost instantly. In that limit, excess returns would exist only in the brief moments between signal discovery and algorithmic exploitation—a market approaching perfect efficiency because machines are doing all the arbitrage.

Whether markets can function effectively in such an environment is an open question. The research provides the mathematical framework for understanding what's happening now but doesn't offer clear guidance on policy responses or how market structure should adapt.

What seems clear is that this is not merely technological tinkering at the margins. The compression of signal half-lives from years to months marks a genuine phase change in market behavior. The implications extend well beyond individual trading strategies to the fundamental economics of active management and the mechanisms by which prices get discovered in the first place.