Citadel Securities Challenges Citrini’s ‘Global Intelligence Crisis’ Thesis

Citadel Securities has issued a pointed rebuttal to Citrini Research’s viral “2028 Global Intelligence Crisis” artificial intelligence (AI) scenario, arguing that current labor data and adoption trends offer little evidence of imminent white-collar collapse.

Flight’s critique zeroes in on what he calls an overconfident leap from technological possibility to economic inevitability. While forecasters routinely struggle to predict payroll growth even two months ahead, he writes, some commentators now claim to see “the forward path of labor destruction” with unusual clarity based on a hypothetical Substack post.

The firm previously described near-term AI capital expenditure dynamics as inflationary, not deflationary. But the heart of this rebuttal lies elsewhere: the speed of diffusion. The displacement narrative, Citadel argues, hinges on the assumption that AI adoption will compound at breakneck speed. Flight states:

“The imminent disintermediation narrative rests on the speed of diffusion.”

So what does the data show? Citing the St. Louis Fed’s Real Time Population Survey, the firm notes that while generative AI usage is growing, frequency data tells a more measured story. If AI were on the verge of replacing broad swaths of labor, daily use for work would likely exhibit a sharp inflection. Instead, the data appears relatively stable.

The Citadel strategist frames the broader debate as a category error: recursive technology does not guarantee recursive adoption. AI systems may improve themselves, but economic deployment historically follows an S-curve. Early uptake is slow and costly, then accelerates as infrastructure matures, before plateauing as integration costs, regulation, and diminishing returns emerge.

Markets, the firm contends, often extrapolate the acceleration phase indefinitely. History suggests otherwise. Organizational change is expensive, regulatory frameworks evolve and marginal gains shrink over time. Slower adoption, in turn, reduces the probability of abrupt displacement.

“Markets often extrapolate the acceleration phase linearly but history implies pace of adoption plateaus as organizational integration is costly, regulation emerges and diminishing marginal returns exist in economic deployment,” Flight’s rebuttal to Citrini’s outlook experiment notes.

Another constraint rarely discussed in dystopian narratives is compute intensity. Training and inference require vast semiconductor capacity, data centers, and energy. Fully automating white-collar work would demand compute at orders of magnitude beyond current utilization. If demand for compute spikes, its marginal cost rises. Should that cost exceed the marginal cost of human labor for certain tasks, substitution stalls. Economic gravity reasserts itself.

Flight also addresses the macro accounting at the core of the Citrini thesis. AI-driven automation is, fundamentally, a productivity shock. Productivity shocks are positive supply shocks: they lower marginal costs and expand potential output. Historically — from steam power to computing — such shifts have raised real incomes over time.

The counterargument claims AI is different because it directly displaces labor income, thereby suppressing demand. Citadel responds with a national income identity: If output rises and real GDP increases, some component of demand — consumption, investment, government spending, or net exports — must also be increasing. A scenario in which productivity climbs while aggregate demand collapses and measured output rises strains accounting logic.

New business formation adds texture to the debate. Data from the U.S. Census Bureau shows a rapid expansion in new business applications. Capital income may have a lower propensity to consume than wage income, but it does not vanish into a black hole. Profits can be reinvested, distributed, taxed or spent.

At the heart of the displacement question lies substitution elasticity — the ease with which firms can replace labor with capital. If that elasticity is extremely high, labor’s share of income could shrink. Yet even then, democratic nations would likely adjust through fiscal and regulatory measures. Moreover, Citadel notes, current labor tracking shows improvement in forward-looking indicators, with AI data center construction contributing to construction hiring.

Flight notes:

“There is little evidence of AI disruption in labor market data as of today. In fact, the forward-looking components of our labor market tracking have improved and AI data center construction appears to be driving a pick-up in construction hiring.”

The economy, Flight argues, consists of countless tasks — physical, relational, regulatory and supervisory — that are costly or difficult to automate. Even cognitive automation faces coordination and liability constraints. It is therefore more plausible, he suggests, that AI will complement labor in many domains rather than eradicate it.

To make his point, Flight invokes John Maynard Keynes’ 1930 essay predicting a 15-hour workweek by the 21st century. Productivity did soar. But instead of withdrawing from labor en masse, societies consumed more. Preferences evolved, new industries formed and human wants proved elastic.

In closing, Citadel sets a high bar for the dystopian scenario to materialize. It would require rapid adoption, near-total labor substitution, no fiscal response, limited investment absorption and unconstrained compute scaling — all at once. Over the past century, technological waves have neither eliminated labor nor produced runaway growth; they have largely sustained long-term trend expansion near 2%.

For Citadel Securities, the AI debate is not about exponential fantasies. It is about substitution elasticities, institutional response and the enduring capacity of human demand to reinvent itself.