Valgo Is Building the Insurance Math for Robots That Do Not Have 30 Billion Claims Records

The Macro: You Cannot Insure What You Cannot Price

Your car insurance premium is calculated from over 30 billion historical claims records. Actuaries at Geico, State Farm, and Progressive have decades of accident data sliced by age, location, vehicle type, driving history, and a hundred other variables. The math works because the data exists.

Autonomous trucks have nearly zero claims records. Warehouse robots have nearly zero. Delivery drones have nearly zero. Every category of physical AI system that needs insurance coverage lacks the historical data that the entire insurance industry relies on to price risk.

This creates a deadlock. Autonomous system operators need insurance to deploy commercially. Insurers will not write policies they cannot price. And the data needed to price those policies will only exist after the systems are deployed at scale. Someone needs to break the cycle.

Valgo, backed by Y Combinator, is the company trying to break it. They build probabilistic risk models from simulation data, turning virtual crash tests, failure scenarios, and environmental conditions into the loss estimates that insurers need to write policies.

The Micro: Stanford PhDs Who Wrote the Textbook

Robert Moss (CEO) and Sydney Katz (CTO) are Stanford PhDs who literally wrote the textbook on validating safety-critical systems and teach the course at Stanford. Jon Qian (President) brings over 12 years of insurance industry leadership and M&A experience. The combination of academic rigor in autonomous systems safety and practical insurance industry knowledge is rare.

The product approach is bottom-up risk modeling. Instead of waiting for historical claims data that does not exist, Valgo builds probabilistic models of routes, tasks, and environments from simulation. They model what can go wrong, how likely each failure mode is, and what the financial impact would be. The output is a simulated loss estimate that an insurer can use to price a policy.

This is technically demanding. You need accurate physics simulation, realistic failure mode enumeration, and statistical methods that produce credible loss distributions. But the founders’ academic work on safety-critical system validation is directly applicable.

The competitive set is thin. Traditional insurance analytics firms like Verisk and LexisNexis Risk Solutions are built around historical data. Newer insurtech companies like Root and Metromile use telematics data from human drivers. Nobody is building simulation-based risk models specifically for autonomous systems at scale.

Valgo operates as a public benefit corporation focused on improving the safety of critical systems. This is a smart structural choice: it signals to regulators and insurers that the company’s models are designed to be conservative and safety-oriented, not to minimize premiums.

The Verdict

Valgo is solving a chicken-and-egg problem that will only get more important as physical AI systems proliferate. Every autonomous truck company, every drone delivery service, every warehouse robotics firm needs insurance, and right now the insurance industry cannot price it.

At 30 days: are insurers actively using Valgo’s simulated loss estimates to underwrite real policies? The gap between “interesting model” and “actuarially accepted” is significant.

At 60 days: how do Valgo’s simulated risk estimates compare to early real-world incident data from autonomous system operators? Model validation against reality is critical.

At 90 days: are autonomous system companies using Valgo’s risk assessments to improve their safety engineering, not just to get insurance? That is the virtuous cycle where better models lead to safer systems.

I think Valgo is positioned perfectly for a market that barely exists today but will be enormous within five years. The team’s academic credibility in safety-critical systems gives them a head start that will be hard to replicate.