ARK Invest’s $24 Trillion Humanoid Robot Scenario

What might the economy look like if humanoid robots became widespread? I review ARK Invest’s September 2024 report, which considers household and manufacturing uses. ARK cites Elon Musk’s view that highly capable humanoids could expand the economy, but its own market figures depend on robots becoming capable and operating at scale.

ARK Invest, founded by Cathie Wood, invests in technologies it considers disruptive and publishes research openly. Its report discusses Tesla’s Optimus, Figure, Chinese robotics companies and Boston Dynamics in the context of a growing industry.

ARK estimates a potential $24 trillion in robot revenues, roughly split between household and manufacturing applications, if humanoids can operate at scale. This is a modeled opportunity, not current revenue or a guaranteed outcome. Possible household tasks include cleaning and dishes; manufacturing examples include factory work.

For the household estimate, ARK combines 2.3 hours of unpaid work per day, 2.8 billion working-age people, a $10.75 weighted average hourly wage and an assumption valuing free time at half of paid time. The resulting modeled opportunity is roughly $12.5 trillion. Its manufacturing estimate varies with productivity uplift and the share captured by robot providers.

ARK says US manufacturing employed nearly 12 million people who worked about 23 billion hours annually, received $785 billion in pay and produced $2.4 trillion in output. These are the report’s baseline figures, not measurements of robot production.

In a deliberately unlikely full-replacement scenario, ARK assumes robots work 80 hours a week for 50 weeks, versus 40 hours for humans. About 5.9 million robots could then supply the same total hours and output as 11.7 million workers, if equivalent productivity is assumed. The report puts robot-provider receipts at roughly $390 billion using a 50% take-rate assumption; this is a hypothetical calculation, not an observed payroll saving.

ARK considers a more likely early scenario in which humanoid robots are expensive and less capable than people. Its net-present-value model compares robot costs against productivity gains: at $16,000 per robot, it estimates that a little over a 5% productivity uplift would make the investment viable under its assumptions.

The chart shows higher robot prices require larger productivity gains. For example, the $40,000 column becomes positive at the 15% uplift row in the displayed model. I find that threshold promising, though whether actual robots achieve it depends on their capabilities and operating costs.

Company size and labor share matter. ARK argues that large manufacturers have more specialized jobs and have benefited from single-task automation, while workers at smaller firms often wear several hats. It says larger firms have historically achieved greater productivity and lower labor cost as a share of revenue, despite paying higher wages.

ARK illustrates this with hypothetical companies. A small firm with $1,000 revenue and 40% labor share assigns $200 per employee; a 25% productivity improvement in the illustration gives a $50 benefit. A larger firm with $5,000 revenue and 20% labor share assigns $250 per employee; a 50% improvement gives $125. These are illustrative assumptions, not measurements of a robot deployment.

ARK argues that small firms may have more tasks that can be automated but have not yet been automated, because existing machines often specialize in one task. A flexible humanoid capable of multiple tasks could therefore be disproportionately useful to them. The report says 40% of US manufacturing employees work for firms with fewer than 500 people.

The report also identifies possible tailwinds: supply-chain shocks after COVID, onshoring and labor shortages. It argues these pressures could make managers less sensitive to robot prices. I see factory closures as an example of why firms might value alternatives, although robots would still require supporting infrastructure and maintenance.

At the start of 2024, I predicted this would be a major year for humanoid robots. By September, I see progress from Tesla Optimus, Figure 2 and Boston Dynamics. I intend to keep following the combination of hardware and AI software as these projects develop.