A recent study by Morgan Stanley's chief economist, Seth B. Carpenter, offers a sobering perspective on the current anxieties surrounding AI. He positions artificial intelligence as the sixth major wave of innovation, following mechanization, electrification, mass production, automation, and the IT revolution, and points out a core paradox: AI is spreading far faster than any previous technological revolution, yet labor market indicators in major global economies exhibit "unusual stability."

From job growth and unemployment rates to job vacancies and turnover rates, these core data points do not show a systematic divergence between industries with high and low AI exposure. Carpenter's research suggests that current evidence leans more towards supporting the assertion that "AI is an incretor rather than a replacement."

Historical lessons: Every technological panic has led to the opposite outcome.

Looking back at the technological leaps since the Industrial Revolution, each one has been accompanied by a deep concern about "machines replacing humans." The Luddites smashing looms in the early 19th century, the fear of automation in the 1960s, and the anxieties about the disappearance of white-collar jobs in the early days of the dot-com bubble in the 1990s have all ultimately proven to be overreactions.

Carpenter's research points out that while these technologies have indeed eliminated some specific tasks and jobs, their more general impact is on changing the composition of work rather than eliminating work itself. Mechanization shifted agricultural labor to factories, electrification spawned a massive service sector, and the IT revolution created entirely new professions such as programmers and data analysts. After each technological leap, the total demand for labor has not shrunk; instead, it has expanded across a broader industrial base.

In my opinion, a frequently overlooked cognitive bias is that many people understand AI as "achieving the same output with fewer people," but the same mechanism also means "the same number of people can create far more output." The two statements are mathematically equivalent, but Morgan Stanley tends to believe the latter is more likely to become a reality. This is due to the aggregate demand expansion effect brought about by increased productivity—when the cost of goods and services decreases, consumers' actual purchasing power increases, thereby generating new demand, which in turn drives employment.

Data shows that productivity gains are driven by output, not layoffs.

Based on the available data, Carpenter believes there is reason for cautious optimism. At the labor market level, indicators such as job growth, unemployment rate, job vacancies, and turnover rate do not show a systematic divergence between industries with high and low AI exposure. Rising youth unemployment is often cited as evidence of AI's impact on employment, but if the cyclical factor of a slowdown in overall US hiring is removed, the excess increase in youth unemployment is only slightly higher than the level predicted by historical cyclical patterns, and does not constitute a structural anomaly.

At the productivity level, the effects of AI are already beginning to show in the data. Industries with high AI exposure are experiencing faster labor productivity growth, but crucially, this growth stems primarily from accelerated output expansion, rather than reduced working hours or staff reductions. This distinction is critical—it indicates that AI is currently playing more of an "incrementer" than a "replacer." Companies are using AI tools to improve the productivity of existing employees, rather than resorting to direct layoffs.

Core risk: The rapid spread of the virus has compressed the adjustment window.

While early data is reassuring, Carpenter clearly points out that future trends remain highly uncertain. Unlike previous technological revolutions that unfolded slowly over decades, the adoption of AI has significantly shortened the adjustment period, which is the most significant structural difference in this wave of innovation.

He raised a cautionary scenario: if companies rapidly realize the productivity gains from AI in the short term, and this effect spreads widely throughout the economy, the unemployment rate could experience a recession-like surge—at least until the labor market has cleared. This "frozen" adjustment would pose a serious challenge to social stability and distributive equity.

However, Carpenter also outlined multiple buffer mechanisms: productivity-driven income growth will support aggregate demand; the rising wealth effect will sustain consumption; new tasks and roles will emerge within firms, absorbing the displaced workforce; the cyclical slowdown in employment and the resulting deflationary pressures will trigger monetary easing; and if monetary policy space is exhausted, fiscal policy's automatic stabilizers and discretionary tools can smooth the income gap during the transition period. He believes that the existence of these buffer mechanisms will make the AI-driven unemployment shock "smaller, shorter-lived, and more controllable."

Infrastructure bottleneck: Over $3 trillion in capital expenditures have yet to materialize.

Carpenter also pointed out that the actual speed of AI's spread will be constrained by the progress of physical infrastructure construction. Morgan Stanley strategists previously predicted that total capital expenditure on data centers and related infrastructure would exceed $3 trillion between 2025 and 2028, but only about a quarter of that has been deployed so far.

This means that AI's greatest impact on productivity and the job market remains largely in the "future tense." The pace of infrastructure construction will directly determine the speed at which AI capabilities penetrate the real economy, thus affecting the window of opportunity for job market adjustments. From chip manufacturing to data center construction, from power grid upgrades to fiber optic cable laying, these physical bottlenecks are becoming "speed limiters" for the implementation of AI.

Policy Response: A Key Variable in Determining the Depth of the Shock

In my opinion, the depth and duration of AI's impact on the job market will largely depend on policy responses. Historically, the adjustment pains brought about by technological revolutions have often been mitigated through education system reforms, improvements to social security networks, and greater flexibility in the labor market. Currently, the challenge facing governments worldwide is: can they establish sufficiently effective retraining systems and social safety nets before AI's penetration accelerates?

From a global perspective, different economies have significantly different policy toolkits. Nordic countries, with their strong union bargaining mechanisms and proactive labor market policies, may find it easier to achieve a smooth transition away from "creative destruction." In contrast, some economies with inadequate labor market protections and weak social security systems may face greater social friction.

Carpenter summarized that Morgan Stanley will continue to track the pace of AI adoption, labor market evolution, and policy responses. "History shows that productivity will ultimately prevail, but not everyone in society will equally share the benefits. Early evidence is encouraging, but the story is still being written." For investors, this means closely monitoring the pace of capital expenditure across the AI ​​industry chain, changes in enterprise adoption rates, and the extent of government intervention in the labor market—these factors will collectively determine the ultimate economic impact of the AI ​​revolution.