Labor's Shrinking Slice: Automation Cuts Workers' Share of US Output

Human labor now accounts for just 8% to 12% of the cost of manufacturing a motor vehicle in the United States, according to the Center for Automotive Research and consulting firm Oliver Wyman, a division of Marsh. That single figure captures the structural shift reshaping the American labor market: output keeps growing, machines do ever more of the work, and a shrinking slice of that output flows to workers as compensation.

The pattern shows up most vividly in semiconductor manufacturing. A new fab may cost $3 billion to build and another $17 billion to equip — over $20 billion in total investment — yet employ only around 2,000 workers when operational. Those jobs include engineers, technicians, supply-chain logistics staff, security and facility management. The output, once the plant runs near capacity, will be enormous.

What does this change for individual workers?

Not motivation — market position. The remaining employees in highly automated facilities are highly trained, well paid and more productive. Real hourly compensation, including benefit costs and adjusted for inflation, has risen sharply over the decades as labor shifted upscale in skills. The 2021–2022 inflation spike, especially in healthcare and other benefit costs, briefly pushed real compensation down, but the long-term trajectory is upward for those who hold these jobs.

The losers are those whose roles vanish entirely. The pools of secretaries taking shorthand on IBM Selectrics no longer exist. Data entry, inventory management, invoicing, payroll and accounts receivable were computerized decades ago. For an individual whose job automation destroys, moving to a higher-skill role, in the source's words, "can prove to be tough or impossible."

The macro picture behind the numbers

The inflation-adjusted value-added output of all workers in nonfarm businesses has doubled since 1999 and quadrupled since 1982, based on the figures Wolf Street cites from national productivity data. Labor makes the output happen, but much of the actual work is done by machines.

Another driver gets less attention: shifting work onto customers. Online banking, ATMs, online brokers, self-checkout, self-service gas stations and automated car washes all create output with little or no labor. Bank labor productivity soared as customers began doing the work themselves.

For employers, this changes the accounting too. Production equipment and machinery count as capital investments, depreciated over time, rather than as ongoing labor costs. That structure rewards further automation:

• Machines: capitalized assets, depreciated over years
• Workers: ongoing compensation costs
• Result: a persistent incentive to shrink labor's share

Is AI different?

The source frames the open question directly: there is "this theory or fear that AI may be the first technology that destroys more jobs than it creates." So far the evidence is mixed. AI has eliminated some jobs while creating others in the AI development ecosystem, the infrastructure buildout, and cybersecurity defending sites such as bank websites against AI hacking. Overall US employment has been rising this year, and the net effect of AI on labor remains undecided.

Historically, technological advances destroyed some jobs but created more than they destroyed — higher-level jobs with more challenging work and better pay, even as labor's share of output declined. Whether AI repeats or breaks that pattern will determine the next chapter of the American labor market.