Benchmarks

Judge It as a Learning Engine

Early deployments are better understood as learning engines than finished products.

Georgian + NewtonX Physical AI Survey, Wave 1 · Reporting base: Buyers, n=85 unless a benchmark states otherwise · July 24 to Aug. 21, 2026

Benchmark H5.1

The Economics Today

Survey question

Thinking about the economics of the physical AI deployments you are closest to, which statement comes closest to your view today?

Audience
All qualified buyers, n=85
In depth
Hypothesis 5 — Judge It as a Learning Engine

Single-select, so the shares add to 100%.

The Economics Today

56%

of robotics buyers say the economics do not work yet, but that they are improving over time

All qualified buyers, n=85

Data for The Economics Today
Category% of buyers
The economics work well today19%
The economics do not work yet, but they are improving over time56%
The economics do not work yet, and they are not improving meaningfully6%
It is too early to assess the economics18%
The economics are not viable for this use case1%
0%20%40%60%The economics work well todayThe economics work well today — % of buyers: 19%19%The economics do not work yet, but they areimproving over timeThe economics do not work yet, but they are improving over time — % of buyers: 56%56%The economics do not work yet, and they arenot improving meaningfullyThe economics do not work yet, and they are not improving meaningfully — % of buyers: 6%6%It is too early to assess the economicsIt is too early to assess the economics — % of buyers: 18%18%The economics are not viable for this usecaseThe economics are not viable for this use case — % of buyers: 1%1%

All qualified buyers, n=85

Benchmark H5.2

Why Buyers Think the Economics Are Improving

Survey question

What makes you confident the economics are improving?

Audience
The 48 buyers who said the economics do not work yet but are improving over time
In depth
Hypothesis 5 — Judge It as a Learning Engine

Select all that apply, so the shares add to more than 100%. Buyers gave 2.4 reasons each on average.

Why Buyers Think the Economics Are Improving

69%

of the 48 buyers who say the economics are improving name less human intervention or supervision per unit of work as a reason for the improvement

The 48 buyers who said the economics do not work yet but are improving over time

Data for Why Buyers Think the Economics Are Improving
Category% of the 48
Less human intervention / supervision needed per unit of work69%
The system handles more of the workflow autonomously over time54%
Each new deployment is faster and cheaper to stand up48%
Support / maintenance burden is falling as the fleet grows42%
Hardware or compute costs are coming down25%
Other0%
0%20%40%60%80%Less human intervention / supervisionneeded per unit of workLess human intervention / supervision needed per unit of work — % of the 48: 69%69%The system handles more of the workflowautonomously over timeThe system handles more of the workflow autonomously over time — % of the 48: 54%54%Each new deployment is faster and cheaperto stand upEach new deployment is faster and cheaper to stand up — % of the 48: 48%48%Support / maintenance burden is falling asthe fleet growsSupport / maintenance burden is falling as the fleet grows — % of the 48: 42%42%Hardware or compute costs are coming downHardware or compute costs are coming down — % of the 48: 25%25%OtherOther — % of the 48: 0%0%

The 48 buyers who said the economics do not work yet but are improving over time

Benchmark H5.3

Expected Payback Period

Survey question

Based on what you’ve seen, what payback period do you expect these deployments to achieve?

Audience
All qualified buyers, n=85
In depth
Hypothesis 5 — Judge It as a Learning Engine

Single-select, so the shares add to 100%.

Expected Payback Period

41%

of buyers expect payback in two to three years. 4% expect it inside a year and 14% beyond three.

All qualified buyers, n=85

Data for Expected Payback Period
Category% of buyers
Under 1 year4%
1–2 years38%
2–3 years41%
More than 3 years14%
We don't expect them to pay back at current economics1%
Too early to tell2%
0%10%20%30%40%50%Under 1 yearUnder 1 year — % of buyers: 4%4%1–2 years1–2 years — % of buyers: 38%38%2–3 years2–3 years — % of buyers: 41%41%More than 3 yearsMore than 3 years — % of buyers: 14%14%We don't expect them to pay back at currenteconomicsWe don't expect them to pay back at current economics — % of buyers: 1%1%Too early to tellToo early to tell — % of buyers: 2%2%

All qualified buyers, n=85

Benchmark H5.4

Pricing Models in Use

Survey question

Which pricing or commercial models are in use in the deployments you are closest to?

Audience
All qualified buyers, n=85
In depth
Hypothesis 5 — Judge It as a Learning Engine

Select all that apply, so the shares add to more than 100%. Buyers named 1.8 models each on average.

Pricing Models in Use

60%

of robotics buyers say outright purchase is in use in the deployments they are closest to, the most common commercial model of the five offered

All qualified buyers, n=85

Data for Pricing Models in Use
Category% of buyers
Outright purchase (capital expense)60%
Subscription / lease (fixed periodic fee)41%
Managed service (vendor operates the robots)36%
Robot-hour or usage-based pricing21%
Outcome-based pricing (pay per unit of work completed)18%
Other0%
None yet1%
0%20%40%60%Outright purchase (capital expense)Outright purchase (capital expense) — % of buyers: 60%60%Subscription / lease (fixed periodic fee)Subscription / lease (fixed periodic fee) — % of buyers: 41%41%Managed service (vendor operates therobots)Managed service (vendor operates the robots) — % of buyers: 36%36%Robot-hour or usage-based pricingRobot-hour or usage-based pricing — % of buyers: 21%21%Outcome-based pricing (pay per unit of workcompleted)Outcome-based pricing (pay per unit of work completed) — % of buyers: 18%18%OtherOther — % of buyers: 0%0%None yetNone yet — % of buyers: 1%1%

All qualified buyers, n=85