Celest Science, a Paris-based climate risk intelligence startup founded in 2023, has closed a €2 million seed round to accelerate the development of its AI-driven forecasting platform. The round was co-led by insurtech-focused venture fund astorya.vc and Silicon Valley accelerator Plug and Play, and supported by 16 business angels drawn from the insurance, finance, and technology sectors. The capital will be directed toward product development and go-to-market expansion across the insurance and financial services industries.
From Proof-of-Concept to Production
The funding follows a structured R&D phase during which Celest Science completed a pilot partnership with global insurer Zurich Insurance Group, transitioning from proof-of-concept to a production-ready deployment. The company has also been selected as one of five startups in Allianz‘s AI accelerator program, an initiative that has backed over 63 startups since its inception, with alumni collectively raising more than €800 million. These early institutional validations signal that Celest’s models are meeting underwriting-grade standards at tier-1 insurers, not just research benchmarks.
Physics-Informed AI Replaces Historical Averages
What distinguishes Celest from established catastrophe modeling firms is the scientific architecture of its models. Traditional risk assessment tools used by the insurance sector rely heavily on historical loss data and statistical averages, a method increasingly inadequate as climate systems depart from past patterns. Celest’s approach integrates deep learning, generative models, and physics-informed AI to analyze interactions across atmospheric, oceanic, and land systems simultaneously.
The platform delivers probabilistic seasonal forecasts ranging from two weeks to six months, directly addressing the temporal blind spot that exists between near-term weather forecasts and multi-decade climate projections. Its models identify the physical drivers behind individual extreme events, correlate compound risk factors, and dynamically adjust frequency curves to account for ongoing climate change trends. The company reports its models are 20% more performant than existing benchmark solutions, though the specific evaluation methodology behind this figure is not independently verified in publicly available documentation.
A Founding Team Bridging Academia and Enterprise
Celest Science was co-founded by Dr. Léo Lemordant and Professor Pierre Gentine. Lemordant previously founded Enerfip, a renewable energy crowdfunding platform that has channeled over €600 million into clean energy projects across Europe. At Celest, he leads commercial strategy and company operations. Gentine heads the “Learning the Earth with Artificial Intelligence and Physics” (LEAP) center at Columbia University and has accumulated more than 500 publications and 20,000 academic citations. His contributions to machine learning applications in climate science were referenced in the 2024 Nobel Prize in Physics committee documentation. Their collaboration began during Lemordant’s doctoral research at Columbia, a background that gives Celest an unusually direct line between frontier climate science and commercial product development.
Investor Thesis: Regulatory Pressure Meets Escalating Losses
The round’s lead investors bring sector-specific rationale. Astorya.vc is a Paris-based venture fund focused exclusively on European seed-stage insurtech, with a portfolio thesis built around emerging risks including climate and cyber. Plug and Play, whose global portfolio includes PayPal, Dropbox, and N26, operates insurtech-focused acceleration tracks across more than 60 international offices and brings both capital and corporate partnership access.
The broader investor group of 16 business angels includes Stéphane Guinet, former AXA Group Executive Committee member and CEO of Kamet Ventures; Christophe Neves, Chief Risk Officer at Skyline and formerly at AIG; Éric Mignot, CEO and co-founder of insurance aggregator +Simple; Didier Valet, Founding Partner at Varsity.vc and former Deputy CEO at Société Générale; and Armando Mann, formerly a senior executive at Dropbox and Google. The composition of the angel syndicate reflects a deliberate positioning at the intersection of institutional insurance knowledge and technology scaling expertise.
Regulatory Tailwinds and a Market Under Pressure
The investment lands in a structurally favorable environment for climate risk data providers. Climate-related economic losses exceeded $100 billion globally in 2023, according to Swiss Re, with France alone recording €6.5 billion in insured losses that year according to industry association France Assureurs. Looking further ahead, Celest’s own data indicates that direct global economic losses from extreme climate events have increased 250% over the past two decades, with projections suggesting a potential doubling of climate-related economic risk by 2050.
On the regulatory side, the EU’s Corporate Sustainability Reporting Directive (CSRD), which entered into force in January 2023 and began phasing in mandatory disclosures from the 2024 financial year onward, requires large companies and financial institutions to formally assess and disclose exposure to physical climate risks. For insurers and reinsurers, this creates a compliance imperative that goes beyond internal risk management and demands auditable, science-backed quantification. Celest is positioning its platform as a response to precisely that requirement.
A Competitive Field With Room for Science-First Differentiation
The climate risk analytics space has attracted growing investment. Comparable ventures include Reask, an Australian platform focused on tropical cyclone risk data for insurers, which closed a $4 million round led by BlueOrchard Finance’s InsuResilience Investment Fund; and Eoliann, a European startup leveraging satellite imagery and machine learning to model flood and climate risk for insurance underwriting. Celest’s differentiation rests primarily on its sub-seasonal to seasonal forecasting capability and the physics-constrained AI architecture, which it argues produces more reliable extrapolation under novel climate conditions than data-only machine learning approaches.
Whether the Zurich partnership and the Allianz accelerator selection translate into recurring commercial contracts at meaningful scale remains the central question as the company moves into its next phase.
