How much does a robot cost instead of an employee, and when does it really pay off?

Investors are spending record amounts on robots meant to replace missing workers on construction sites, in factories and warehouses. We looked into how much such a robot costs compared with a full-time position and where these calculations stand up to real-world conditions.

Money flows where people are missing

Companies combining construction with robotics raised about $1.6 billion this year, as of October 29. According to Crunchbase, that is more than four times as much as in all of 2025 ($389 million) and three times as much as in the previous record year of 2021 ($506 million). The number of rounds hardly changed: 30 versus 28 a year earlier. So the money is not going to a larger number of companies, but in larger amounts to just a few of them. Bedrock Robotics, Icon and Gravis Robotics together raised about 83 percent of the entire pool.

The reason is straightforward. In this year's survey by the American contractors' association AGC and the training organization NCCER, 42 percent of companies cited labor shortages as a cause of delays. It was the most frequently mentioned cause, and 87 percent of companies have open vacancies in manual roles. The boom in building data centers for AI is intensifying competition for workers even further: 58 percent of companies that worked on them felt stronger competition for employees. Industry is similar. Deloitte and The Manufacturing Institute estimate that by 2033, 1.9 million jobs in American factories may remain unfilled.

A humanoid on subscription, or the first price list on the table

The most hard numbers came with the debut of Agility Robotics. The maker of the Digit humanoid entered Nasdaq on November 5 through a merger with a SPAC, a special purpose company set up to bring another company to the stock exchange. As the first such manufacturer in the US, it had to present its business assumptions in filings for the American securities regulator.

In the RaaS rental model (robot as a service), the customer pays about $25,000 one-off for delivery and deployment. On top of that comes $8,500 per month for software and service, or about $102,000 per year. With a purchase, the customer pays $200,000 for the hardware, $20,000 for deployment, and $36,000 per year for software and service. The robot is supposed to operate for about five years and work more than 20 hours a day: nine minutes of battery charging provides 90 minutes of work.

Chief financial officer Michael Beer told analysts that renting replaces $200,000 in annual labor costs at a cost of about $100,000. Purchasing is meant to replace $1.1 million in wages over five years at a cost of about $400,000. The second figure matches the price list: $220,000 upfront plus five years of fees at $36,000 gives exactly $400,000.

The catch lies elsewhere. According to our calculations, $200,000 a year is not one salary, but work across several shifts. For comparison, the median pay of a construction equipment operator in the US is about $59,600 a year according to the BLS labor statistics office, and that is without benefits and overtime, so less than the full cost to the employer. So the robot wins where it works almost around the clock. On a single shift, the result is completely different.

Does it work? Humanoids are only now being put to the test

Agility itself admits in the prospectus that humanoids have not been tested at large scale. In 2025, the company had $1.8 million in revenue and a $140 million operating loss. Its order backlog exceeds $300 million. The core of it is a contract from an undisclosed customer for 1,000 robots under a three-year subscription. That works out to about $100,000 per robot per year, which is exactly what follows from the price list. The money will only come in, however, if Agility meets the contractual deadlines and product parameters.

The new Digit 5 will not be available until 2027 and will lift up to about 22.7 kg. Earlier versions of Digit are already working for several customers: in GXO warehouses, at Schaeffler's factory and at Toyota in Canada. Investors were cautious, however. About 61.5 percent of SPAC shareholders requested their money back, reducing the funds in the special purpose company's trust account from about $424 million to $164 million. An analyst at The Motley Fool advises against treating Agility's long-term projections as a binding forecast and instead waiting for results from a few quarters. Hurst himself tempers expectations for applications outside the factory floor:

Excavators and carts: a more mature part of the market, but without price lists

Further along are robots that do not imitate humans, but take over one machine or one task. Bedrock Robotics converts existing excavators into autonomous ones and raised $870 million this year at a valuation of $4.1 billion. According to the company, its excavators have been working without operators on commercial construction sites in Texas and Nevada since August. Bedrock does not publish a price list. All that is known is that the excavator itself costs from $0.5 million to $1.5 million. Analysts therefore advise comparing the cost of moving a unit of earth, rather than the operator's rate. Downtime, fuel, rework and equipment availability can matter more than wages.

It is similar in factories. Aformic from Gliwice, spun off from industrial automation company AIUT, raised $6.95 million (about PLN 27 million) in a round led by US fund Veteran Ventures Capital, with participation from ff Venture Capital among others. Its autonomous mobile robots (AMRs) transport parts between the warehouse and the production line for customers such as Toyota, Hitachi and 3M. According to the company, over the course of a year they covered nearly 300,000 km and handle more than 5,000 orders per day. Industry outlet Tectonic reports that in 2025, Aformic had about $10 million in revenue. The company does not disclose the prices of its robots. Industry estimates point to a typical return on an AMR deployment after 18-24 months, but these are advisory calculations, not audited data. In the same week, Tutor Intelligence raised $75 million from fund a16z (Andreessen Horowitz) for robots rented to factories by the hour.

How to calculate a robot in your own company

Our conclusion: the "robot instead of a full-time position" equation can already work out, but under narrow conditions. The numbers from Agility's prospectus and manufacturers' declarations are a starting point, not proof. Before a company starts talking to a supplier, it is worth checking four things:

• whether the task is repetitive and takes place in an organized environment,
• whether the robot will work two or three shifts rather than one,
• whether we are comparing the full labor cost, with benefits and downtime, rather than just salary,
• whether renting with the option to cancel allows the company to test the robot without buying equipment.

Today, the best-documented robots are transport robots in factories and converted construction machines. Humanoids have their first public price list, but also operating losses of around $140 million a year and a new model not available until 2027. Anyone who has trouble staffing shifts today should first take a closer look at simpler robots.