Authored by global macro analyst Frank Flight, the market-making firm’s response to Citrini Research’s memo begins with present-day facts rather than forward fiction. As of 2026, unemployment sits at 4.28%, AI capital expenditures amount to roughly 2% of GDP — about $650 billion — and nearly 2,800 data centers are planned across the U.S., according to the firm’s figures. Meanwhile, job postings for software engineers are up 11% year over year.

For readers unfamiliar with Citadel Securities, the company is one of the largest liquidity providers in global markets, active across equities, options, and fixed income. It operates separately from the hedge fund Citadel, though both share historical ties to founder Ken Griffin. When Citadel Securities speaks, it does so from a vantage point steeped in macro data and market plumbing.

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.

• What did Citadel Securities argue in its rebuttal?
  The firm contends that current labor data and AI adoption trends do not support imminent mass displacement of white-collar workers.
• Who is Citadel Securities?
  It is one of the largest global market makers, providing liquidity across equities, options, and fixed income markets.
• Does Citadel believe AI is deflationary or inflationary?
  The firm has said near-term AI capital expenditure dynamics appear inflationary rather than contractionary.
• What is substitution elasticity in the AI debate?
  It refers to how easily firms can replace human labor with AI capital without significant cost increases.

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。