AWS study puts Israel joint second for physical AI startup share across 20 markets
Germany leads the surveyed markets, but the study’s bigger constraint is demand: 70% of physical AI builders say slow customer adoption limits growth.
Loading page…
Germany leads the surveyed markets, but the study’s bigger constraint is demand: 70% of physical AI builders say slow customer adoption limits growth.
Listen to this story
AWS’s October 2026 index finds that physical AI builders make up 21% of surveyed startups in Israel, tying it with Japan; Germany leads at 24%. The ranking measures each market’s share of surveyed startups working in physical AI, not company counts or revenue. The bigger commercial hurdle is deployment: 70% of builders say slow customer adoption constrains growth, while AWS attributes the gap partly to customers not seeing the technology in action. Its proposed pilots and demonstrations are recommendations, not proven remedies.
The index covers 4,420 startups across 20 countries, including 663 physical AI builders; 15% of the full sample qualifies as builders.
Physical AI work is broader than humanoids: 55% of builders work on computer vision or perception, while 12% work on humanoid or fully autonomous robots.
Builders rate local labor and skills conditions at 68%, supply chains and components at 60%, capital at 46%, and regulation at 40%.
Israel has one of the highest concentrations of startups building AI for the physical world among 20 markets surveyed by AWS. At 21%, it ties Japan behind Germany’s 24%, CTech reported November 16. But the study identifies a constraint beyond building machines: 70% of physical AI builders say slow customer adoption limits growth.
The country comparison measures the share of surveyed startups working in physical AI, not total company numbers or revenue. The United States stands at 20%, just below Israel and Japan. Across the full sample, 15% of startups qualify as physical AI builders.
AWS’s AI Gets Physical: The Global Startups Readiness Index is dated October 2026. It draws on 4,420 technology startups across 20 countries, including 663 physical AI builders. Its readiness measures capture builders’ assessments of local conditions, rather than an independent test of how well their machines perform.
Physical AI connects artificial intelligence to systems that sense their surroundings and act in the real world. That includes robots, drones, vehicles and industrial equipment. AWS uses a broad definition of builders: companies making hardware, embedding physical AI in products, or developing the software and technologies that enable those systems.
The resulting picture is less humanoid-heavy than the label might suggest. Computer vision and perception—software that helps machines interpret their surroundings—feature in 55% of builders’ work. Sensors and chips embedded in physical products feature in 50%. Only 12% work on fully autonomous or humanoid robots.
For many respondents, AI is not a side feature. Some 67% of physical AI builders describe AI as their entire proposition or core product, compared with 47% of all surveyed startups. For 49% of builders, physical AI itself is the core product or the whole business.
Market readiness is the weakest of the index’s five conditions for scaling. Labor and skills score highest at 68%, followed by supply chains and components at 60%, capital and investment at 46%, and regulation at 40%. Each score reflects the share of builders who say their market provides the relevant conditions.
AWS argues that limited awareness, rather than a lack of appetite, explains much of the adoption gap. Many potential customers have not seen physical AI working or understood how to apply it. The report recommends pilots, live demonstrations and industry events to make those uses visible. That is AWS’s diagnosis and proposed remedy, not a measured result from those interventions.
There is already demand within the startup sample. Sixteen percent of all respondents use physical AI systems built by others, 12% are piloting them, and 29% are evaluating adoption. Those figures describe startup buyers, not the broader business population whose readiness builders were asked to assess.
Customer adoption is not the only obstacle. Physical AI requires hardware to be designed, built and tested in real-world conditions, making development more expensive and time-consuming than software. The survey identifies several pressures on that process:
Manufacturing and automotive offer a concrete view of deployment today. They are the industries builders most expect physical AI to transform first, cited by 55%. Among builders serving that sector, 51% say their products are already commercially deployed—putting factory-floor uses beyond the prototype stage for roughly half of that group.
Loading discussion...
Join the conversation
Explain when caution is sensible—and when it becomes a missed opportunity.
Be the first to share a perspective or an experience.
Reader comments
Newest comments first. Replies stay oldest first.