McKinsey predicts physical AI value will shift from hardware to deployment
The firm forecasts over $1 trillion in physical AI value by 2035 as the focus moves toward software, field data, and operational expertise.
The short version
- McKinsey moved its $1 trillion physical AI value milestone forward from 2040 to 2035.
- Current robotics funding is lopsided, with $33 billion of $38 billion going to makers and model developers.
- Future value is expected to reside in deployment partners, field data, and specialized components like tactile sensors.
- The most critical training data will likely be records of robots failing and humans stepping in to correct them.
- Business models are shifting toward 'robots-as-a-service' priced by task or outcome rather than hardware sales.
Why it matters
For businesses, this means the most important part of a robotics deal may soon be the quality of the deployment data and the integrator's skill rather than the robot's specs. It suggests that the 'intelligence' of the system becomes a commodity, while the ability to run it reliably in a specific environment becomes the primary competitive advantage.
A shift in economic value
McKinsey has updated its economic outlook for physical AI, now predicting that these systems will generate over $1 trillion in value by 2035. This is a significant acceleration from a previous estimate that placed the $1 trillion milestone in 2040. The firm defines physical AI as systems using sensors and learned models to act in the physical world with minimal human help.
While the technology is currently in a high-growth phase, the nature of the profit is expected to change. Currently, most investment flows toward the companies building the robots and the underlying foundation models. However, McKinsey expects the center of gravity to move toward the software, deployment strategies, and the actual field operations where the robots work.
This shift suggests that the most valuable assets in the coming years will not be the physical machines themselves. Instead, the value will reside in the site-specific data and the expertise required to integrate these machines into complex environments. As the technology matures, the buyer's own floor and the data generated there become the primary drivers of value.
The data loop of failure
A key driver of this transition is how robots learn from their environments. While simulations are useful for basic navigation, they struggle to replicate the nuances of friction, force, and slip. These small physical details are best learned through live operations, making field deployment the richest source of insight for developers.
The most valuable training data may actually come from mistakes. When a robot fails to complete a task and a human has to step in to intervene, that specific interaction provides a high-quality data point for the model. Every pilot program on a customer's floor essentially functions as a data collection run for the vendor, improving the system for future use.
Because of this, the terms of these contracts are becoming as important as the hardware's uptime. Companies need to scrutinize how deployment data and intervention logs are handled by vendors. If a vendor uses a customer's failure data to improve their general models, that data becomes a significant piece of intellectual property.
Current funding vs. future value
There is currently a mismatch between where the money is going and where McKinsey expects the value to land. Of the $38 billion invested in general-purpose robotics, about $33 billion went to robot makers and model developers. In contrast, deployment partners and key components like actuators and tactile sensors shared less than $6 billion.
Real-world scale and growth
The transition toward physical AI is already visible in large-scale industrial applications. For example, Caterpillar operated a fleet of 690 autonomous trucks at the end of 2024 and has set a goal to triple that number by 2030. This demonstrates how structured environments, like mines, are providing the necessary repetition for systems to gain experience.
The investment landscape is also moving rapidly, with funding for general-purpose robotics showing a steep climb. In the first eight months of 2026 alone, the total funding reached a level that is already more than double the entire figure recorded for all of 2025. This rapid influx of capital highlights the urgency to move from pilot programs to full-scale deployment.
Separating body from brain
While many current developers are vertically integrated to control both sensors and policies, new designs are beginning to decouple hardware from intelligence. In June 2026, NVIDIA released an open reference design for humanoids that explicitly separates the body from the brain. This approach allows for a more modular way to combine different chassis and tactile hands with specific compute models.
Despite these architectural shifts, the market for humanoids is still in its early stages. By the first half of 2026, roughly 19,100 humanoids had shipped worldwide. As these numbers grow, the focus will likely shift toward how these different components can be integrated into specific tasks, moving away from the need for a single company to build every part of the machine.
Sources and links
- ReportingMarketScale: the article we sourced
Read the original at MarketScale
First reported by MarketScale. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.