CLAI Ventures Invests in Vayuh: AI Foundation Models for Extreme Weather Risk
In the U.S. alone, severe convective storms (hail, tornadoes, and straight-line winds) now account for $54 billion in annual losses, overtaking hurricanes as the top driver of catastrophe payouts. Gallagher Re has published numbers for global payouts in the first half of 2025. Insurance-covered losses from natural catastrophes are rising rapidly. In the first half of the year alone, insured losses reached at least $84 billion, which is 55% higher than the 10-year average of $54 billion. Total economic losses climbed to $151 billion, also above the decadal average. Much of this surge was driven by costly U.S. severe convective storms and major wildfires in the Los Angeles region, clear signs that extreme weather events are increasingly translating into larger insured payouts.
Yet, the core models that insurers depend on to price and transfer risk mainly assume that future patterns will mirror historical patterns. Traditional catastrophe models rely on historical records and physics-based equations, both of which struggle to predict hyper-local and tail events that now dominate insurance losses. To compensate for this uncertainty, insurers routinely add 20–50% “safety loads” on top of their model outputs, eroding profitability and distorting capital allocation.
Carriers continue to seek new solutions to manage this new dynamic risk, and contain this volatile loss for climate-exposed regions. At the same time, of course, the business continues to seek new insurance products to profitably write.
The Industry Landscape
A wave of innovation is reshaping how risk is measured, priced, and underwritten. AI-native insurers like Stand Insurance, which recently raised a $35 million Series B, are proving that machine learning can make weather-driven insurance viable and scalable. Yet the underlying infrastructure for modeling natural catastrophes remains stuck in the past, dominated by legacy providers such as Moody’s and RMS, whose systems were designed for an era of slower, more predictable climate dynamics.
The global Property & Casualty insurance market exceeds $1 trillion annually, with natural catastrophes responsible for up to 20% of claims. This represents an enormous opportunity for data-driven disruption. As parametric and API-based insurance products emerge, the demand for faster, more accurate, and globally adaptable catastrophe models is skyrocketing.
What cloud infrastructure did for computing, AI foundation models for catastrophes will do for risk, powering a new generation of insurers, reinsurers, and brokers who need to quantify the unquantifiable.
The Unique Founders Behind Vayuh
Dr. Mayur Mudigonda, founder of Vayuh, met Dr. Pratik Sachdeva at UC Berkeley, where they earned PhDs in theoretical neuroscience and worked on the earliest applications of AI to weather and climate. Pratik now leads much of Vayuh’s science work. Mayur’s research at Lawrence Berkeley National Lab pioneered AI-based wildfire modeling and earned his team the ACM Gordon Bell Prize, one of the highest honors in high-performance computing.
Unlike conventional models that simulate weather forward in time, Vayuh’s approach is stochastic and retrospective. By learning the physics of catastrophes directly from diverse datasets (radar, satellite, reanalysis, and insurance portfolios) their models generate detailed 100,000+ year event catalogs that help insurers more accurately price, manage, and transfer risk.
Critically, this modeling approach leverages AI’s true power: fast inference. This enables Vayuh to create accurate hazard layers for regions with little to no historical data, including emerging markets like Canada, Mexico, Australia, and EU, unlocking new geographies for insurers that legacy models struggle to cover.
The result is a 10–20% accuracy improvement and 100–1000x faster inference compared to traditional physics-based models. Additionally, the advanced AI models enable a more detailed and accurate inference, generating updated risk assessments annually compared to the traditional decade-long intervals. These gains are not theoretical; early clients in the insurance space are already leveraging Vayuh’s models to optimize line sizing, portfolio rebalancing, and reinsurance strategies.
Why We Are Excited to Back Them
At CLAI Ventures, we invest at the intersection of AI and resilience, technologies that help humanity adapt to accelerating environmental and systemic volatility. Vayuh embodies this vision.
We were drawn to their technical depth, measured commercial discipline, and clear product-market timing. This seed round offers an attractive entry point given the caliber of the founding team and the sophistication of their first customers. We also see powerful parallels between Vayuh and early infrastructure companies in other sectors: just as Databricks became the data layer for enterprise AI, Vayuh has the potential to become the risk modeling layer for the insurance industry.
As insurers, reinsurers, and financial institutions race to quantify extreme weather risk, Vayuh’s foundation models will power the transition from reactive to proactive risk management. The company’s ability to scale across geographies, from the U.S. to Mexico, Europe, and Australia, further strengthens its positioning as a global leader in this emerging category.
We’re thrilled to support Mayur, Pratik, and the Vayuh team as they build the AI infrastructure for a world defined by climate volatility, helping insurers and communities alike better understand, price, and withstand the storms ahead.
This post was written by Ajay Gupta and the CLAI Ventures team, a Silicon Valley-based fund investing at the intersection of AI and climate. You can reach me at ajay@clai.vc.