Anthropic / Wikimedia Commons (Public domain)
Anthropic report finds robots can handle 74% of US physical tasks, but cost is the catch
Robots are technically capable of most physical work but cheaper than humans for only 0.3% of job tasks, according to Anthropic
Robots can technically perform about three-quarters of the physical tasks done by US workers. They are cheaper than a human for just 0.3% of them.
That gap sits at the center of a new Anthropic report. It reframes the robots-are-coming-for-your-job debate as a question about price, not ability.
The report, titled “What work can robots do?”, was published on September 30, 2026. It introduces a robot exposure index, a scoring system that measures what current robotic technology can actually do across nearly 900 US occupations. The analysis draws on occupational data from O*NET to break those jobs into individual tasks.
The headline figure: robots can handle 74% of physical job tasks in the US. That share represents approximately 34% of total working hours across the economy.
Capable is not the same as affordable
First, it covers only physical work. According to the report, physical tasks make up 46% of working time overall.
Second, the robotic capability Anthropic measured is largely confined to structured or purpose-built environments.
Third, and most important, is cost. Robots are currently competitive with human labor in only 0.3% of job tasks.
The report translates that into about 300,000 workers who could be fully replaced by automation today.
The timeline for closing the gap is long. At historical rates of price decline, the cost-competitive share would only reach 10% in approximately 40 years.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Getting to that 10% mark would require a 70% reduction in robot costs.
Who is most exposed
The exposure is not spread evenly. According to the report, the workers most exposed to robotic automation are predominantly male, less educated, and lower-paid.
Vehicle operators dominate the list. Nine of the ten most exposed occupations fall into that category.
Taxi drivers rank highest, with an exposure index score of 2.2. Warehouse work also appears among the most affected occupations.
The wage split is stark. Highly exposed workers earn around $23 per hour, according to the report. Unexposed workers earn about $53 per hour.
On the other end, roles that demand advanced interpersonal skills or unstructured physical ability look safer for now. The report points to nursing and maintenance as examples of work less likely to be automated in the near future.
Why the capability-versus-cost framing matters
Anthropic’s report shifts attention to whether deploying a robot actually makes economic sense. A machine that can do a task but costs more than the person doing it is a science project, not a replacement.
Physical tasks account for less than half of working time. The remaining cognitive and interpersonal work is where robots are weakest.
What this means for investors and workers
The report’s 40-year projection rests on historical rates of price decline. If costs fall faster than that historical pace, the timeline would compress.
Because exposure clusters among lower-wage workers, the effects of automation would fall unevenly when they do arrive.
Vehicle operators deserve particular attention. With nine of the ten most exposed occupations in that category, transportation is the sector where the gap between capability and cost would matter most if it narrows.
For now, the 300,000 workers the report identifies as cost-replaceable represent a small slice of the economy. The more meaningful number may be the 74%: it marks how much physical work is already within technical reach, waiting on price.