JPMorgan Chase: AI is not a job stealer, but a productivity multiplier; demand expansion is the key to employment.

Morgan Stanley points out that the speed of AI diffusion far exceeds any historical technological revolution, yet the labor market still demonstrates extraordinary stability; AI is currently more an incrementer than a replacer.
(Background: Who said AI coins are FET? The true machine economy’s winner is actually USDC.)
(Additional context: Morgan: Trade wars could cause TSMC and other tech stocks to plummet 20%, suggesting profit-taking first.)

Morgan Stanley Chief Economist Seth B. Carpenter’s latest research provides a wake-up call amid the collective anxiety surrounding AI. He positions artificial intelligence as the sixth major wave of innovation following mechanization, electrification, mass production, automation, and the IT revolution, and highlights a core contradiction: AI’s diffusion speed far surpasses any previous technological revolution, yet key indicators in major economies’ labor markets show “unusual stability.”

From employment growth, unemployment rates, to job openings and resignation rates, these core data points do not show systematic divergence between high-exposure and low-exposure AI industries. Carpenter’s research argues that current evidence better supports the view that “AI is an incrementer, not a replacer.”

Historical Lessons: Every Tech Panic Ended in the Opposite

Looking back at technological leaps since the Industrial Revolution, each wave has been accompanied by deep fears of “machines replacing humans.” Early 19th-century Luddites destroyed weaving machines, 1960s automation fears, and early 1990s concerns over white-collar job losses during the dot-com bubble—all proved to be overreactions, as history has shown.

Structural Insights: Tech Shifts Reshape, Not Eliminate, Jobs

Carpenter emphasizes in his report that these technologies have indeed replaced certain specific tasks and roles, but more generally, they have reshaped the composition of work rather than eliminated jobs altogether. Mechanization shifted agricultural labor into factories, electrification gave rise to large service sectors, and the IT revolution created new professions like programmers and data analysts. After each technological leap, labor demand did not shrink; instead, it expanded across broader industries.

He notes an often-overlooked cognitive bias: many people interpret AI as “fewer people doing the same output,” but the same mechanism also means “the same number of people can produce much more.” Mathematically equivalent, but Morgan leans toward believing the latter is more likely in reality. This is driven by productivity gains leading to increased overall demand—when the costs of goods and services fall, consumers’ real purchasing power rises, creating new demand and thus boosting employment.

Empirical Data: Productivity Gains Driven by Output, Not Layoffs

Based on current data, Carpenter believes there is reason for cautious optimism. In labor markets, indicators like employment growth, unemployment, job openings, and resignation rates show no systematic divergence between high- and low-exposure industries to AI. The rising youth unemployment rate is often cited as evidence of AI’s impact, but when cyclical factors in overall US hiring slowdown are excluded, the excess increase in youth unemployment is only slightly above what historical cycles predict, not a structural anomaly.

Output Expansion Comes First: AI Increases Capacity, Not Headcount Cuts

On productivity, data already shows effects. Labor productivity growth is faster in high AI exposure industries, but the key point is that this growth mainly results from accelerated output expansion, not from reduced working hours or layoffs. This distinction is crucial—it indicates AI currently plays a role more as an “incrementer” than a “replacer.” Companies are using AI tools to boost existing employees’ productivity, not directly firing workers.

Core Risk: Rapid Diffusion Compresses Adjustment Window

While early data is reassuring, Carpenter explicitly states that future trends remain highly uncertain. Unlike past technological revolutions that unfolded over decades, AI adoption has significantly compressed the adjustment cycle—this is the most notable structural difference in this wave of innovation.

He warns of a scenario to watch closely: if companies quickly realize productivity gains from AI in the short term and this effect broadly spreads through the economy, unemployment could spike similarly to a recession—at least until the labor market clears. This “fast-freeze” adjustment could pose serious challenges to social stability and income distribution.

Buffer Mechanisms: Can Six Lines of Defense Absorb the Shock?

However, Carpenter also lists multiple buffer mechanisms: productivity-driven income growth will support aggregate demand; wealth effects will sustain consumption; firms will create new tasks and roles to absorb displaced workers; cyclical slowing in employment and associated deflationary pressures will trigger accommodative monetary policy; if monetary policy space is exhausted, automatic stabilizers (like unemployment benefits, progressive taxes) and discretionary fiscal tools can help smooth the cycle. He argues that these buffers will make the employment shocks from AI “smaller, shorter, and more controllable.”

Infrastructure Bottlenecks: Over $3 Trillion Capital Expenditure Not Yet Realized

Carpenter also notes that the actual diffusion speed of AI will be constrained by physical infrastructure development. Morgan’s strategists previously forecast that between 2025 and 2028, capital expenditures on data centers and related infrastructure will surpass $3 trillion, but only about a quarter of that funding has been secured so far.

Hardware Bottlenecks Determine Penetration Speed: Chips, Power Grids, and Fiber Are Speed Limits

This means that AI’s greatest impact on productivity and employment remains “future-oriented.” Infrastructure development pace will directly influence how quickly AI capabilities penetrate the real economy, affecting the timing of labor market adjustments. From chip manufacturing to data center construction, from grid upgrades to fiber optic deployment, these physical bottlenecks are becoming the “speed limiters” for AI deployment.

Policy Responses: Key Variables in Impact Magnitude

Carpenter emphasizes that the depth and duration of AI’s impact on employment will largely depend on policy responses. Historically, adjustment pains from technological revolutions have been alleviated through reforms in education, social safety nets, and labor market flexibility. The key challenge for governments now is whether they can establish effective retraining systems and social protections before AI’s rapid penetration.

From a global perspective, policy toolkits vary significantly across economies. Nordic countries with strong union bargaining and active labor policies may transition more smoothly through “creative destruction,” while economies with weak social safety nets and labor protections could face greater social friction.

Carpenter concludes that Morgan will continue monitoring AI diffusion speed, labor market evolution, and policy responses. “History shows productivity will ultimately prevail, but not everyone in society will share the benefits equally. Early signs are encouraging, but the story is still being written.” For investors, this means closely watching capital expenditure trends across AI supply chains, corporate adoption rates, and policy measures affecting the labor market—these factors will jointly determine the ultimate economic trajectory of the AI revolution.