Hypothesis 3: A Systems Market, Not a Model Market

What we said

Physical AI is not simply a model market. It requires a broader stack, including hardware, data, evaluation, observability, safety, deployment operations and outcome systems alongside the model itself. If that is right, the people running deployments should tell us the surrounding work is where both the difficulty and the money sit.

How we tested it

We looked at five measures:

What buyers name as the hardest part of making physical AI work.
What stops them scaling.
How they split the cost of a deployment.
Where the surrounding spend goes.
How they obtain the capability in the first place.

What we found

Buyers describe physical AI as a systems problem and buy it as a hardware-and-model purchase. Asked what is hardest, they name fitting the robot into a plant that already runs rather than the autonomy model. Asked what stops them scaling, they say much the same. But asked to divide the cost of a deployment, they put roughly 70% on the robot hardware and the underlying model and 30% on everything around them (data, evaluation, operations, integration, safety and support). That 30% goes mostly on running the system and fitting it to the workflow.

The Hardest Part of Making Physical AI Work

69%

of robotics buyers rate integrating the robot into their existing workflow among the three hardest parts of making physical AI work. It is the highest of any option, and the most-named first choice at 32%.

All qualified buyers, n=85 · Buyers ranked their top three, so the shares add to more than 100%

Data for The Hardest Part of Making Physical AI Work
Category	In top three (%)
Integrating it into the real workflow / environment	69%
Operating it live — monitoring, diagnosing failures, updating safely	55%
Getting enough good data and being able to simulate / test	45%
Knowing how to measure whether it's actually working (evaluation)	34%
Safety and security	32%
The underlying AI model / autonomy	32%
The robot hardware itself / on-device compute	21%
Still evaluating / not applicable	12%

All qualified buyers, n=85 · Buyers ranked their top three, so the shares add to more than 100%

See the full benchmark

What Stops Buyers Scaling

51%

of robotics buyers name integrating with existing operations and systems among their three biggest barriers to deploying or scaling. It is the most-named barrier.

All qualified buyers, n=85 · Buyers ranked their top three, so the shares add to more than 100% · A “no significant barriers” option was offered and taken by nobody

Data for What Stops Buyers Scaling
Category	In top three (%)
Integration with existing operations and systems	51%
Upfront cost / capital required	44%
Reliability / performance in real conditions	39%
Security and data concerns	28%
Difficulty changing existing workflows / processes	24%
Unclear or unproven ROI	22%
Workforce concerns and change management	21%
Availability of high-quality training data	20%
Supply chain or manufacturing capacity constraints	19%
Safety and risk to people	14%
Lack of internal expertise	13%
Vendor dependence / lock-in	6%

All qualified buyers, n=85 · Buyers ranked their top three, so the shares add to more than 100% · A “no significant barriers” option was offered and taken by nobody

See the full benchmark

Where Deployment Cost Goes

70%

of deployment cost goes to the robot and the AI model per robotics buyers’ own estimates, leaving 30% for the integration, evaluation, operations, safety and support work that surrounds them

Qualified buyers who allocated, n=77 — the other 8 answered “still evaluating” and are excluded rather than counted as zero · Spread across buyers: robot hardware 10–80%, AI model & autonomy 0–70%, everything around them 5–80%

Data for Where Deployment Cost Goes
Category	Mean	Median
Robot hardware	41%	40%
The AI model & autonomy	28%	25%
Everything around them (data, evaluation, operations, integration, safety, support)	30%	30%

MeanMedian

See the full benchmark

Where the Surrounding Spend Goes

70%

of the robotics buyers who spend meaningfully on the work around the robot put the largest share of that spend into live operations (monitoring, fleet management, incident response, updates) or workflow integration

Qualified buyers who put at least 30 of their 100 points on the surrounding category, n=43 · single-select, so the shares add to 100%

Data for Where the Surrounding Spend Goes
Category	%
Live operations (monitoring, fleet management, incident response, updates)	37%
Workflow / environment integration and customization	33%
Data collection, simulation and evaluation	16%
Safety, security and compliance	9%
Field support and human supervision / teleoperation	5%

Qualified buyers who put at least 30 of their 100 points on the surrounding category, n=43 · single-select, so the shares add to 100%

See the full benchmark

Directional

How Buyers Obtain the Capability

71%

of robotics buyers take a complete solution from a robotics vendor — but buyers already running at scale add their own assembly on top, with 56% also using a systems integrator and 53% buying hardware separately

All qualified buyers, n=85 · Select all that apply, so the shares add to more than 100% · Segment bases are small (sizes shown in each column header); read differences across segments as directional, not precise

Data for How Buyers Obtain the Capability
Category	Scaled (n=32)	Limited production (n=29)	Piloting (n=24)
Complete solution from a robotics vendor (hardware + software)	69%	66%	79%
Hardware bought and integrated / configured separately	53%	59%	21%
Through a systems integrator	56%	38%	25%
Built in-house	25%	3%	25%
Rented / robots-as-a-service	19%	17%	8%
Other	0%	0%	4%

Scaled (n=32)Limited production (n=29)Piloting (n=24)

See the full benchmark