Kettle
California-focused climate reinsurance MGA that uses convolutional neural networks to model wildfire risk at half-square-mile resolution, then writes wildfire reinsurance and commercial E&S property coverage underwritten by Mt. Hawley Insurance Company (RLI Corp.), an A+-rated fronting carrier.
ourkettle.com ↗Score
- Traction (named carrier deployments)2 carrier deployment(s) with public source.
- 1/5
- Maturity (years since founding)6 years since founding (2020).
- 2/5
- Coverage (insurance lines supported)3 line(s) supported: reinsurance, commercial, specialty.
- 3/5
- Analyst recognition (Celent / Gartner / Forrester / Everest / ISG)6 mention(s).
- 2/5
What it does
Kettle launched in early 2020 as Kettle Re, a San Francisco-based reinsurance MGA (a managing general agent — a company licensed to underwrite and bind reinsurance on behalf of capacity providers) built around the idea that wildfire risk in California had become too complex for models trained on historical loss data. The founders are Andrew Engler, formerly VP of Digital at Argo Group, and Nathaniel Manning, who had been CEO of Ushahidi and the first chief data officer of USAID. Son Le, the third co-founder, remains at the company as CTO.
The model. Kettle's core technology is a set of convolutional neural networks that run more than 140 million parameters across 130 terabytes of data drawn from over 35 sources: NOAA weather records, NASA MODIS and LiDAR satellite imagery, state real estate databases, topography, vegetation, and building characteristics. The model runs roughly 2.25 million fire-spread simulations per cycle — compared with the approximately 10,000 simulations used by conventional probabilistic cat models. Output is a property-level probability-of-burn score at half-square-mile resolution across California. Kettle claims the model predicted 14 of California's 14 largest fires in 2020 as being in the top 20% of riskiest geographic areas.
Funding and backers. Kettle raised $4.71 million in seed funding in October 2020, led by True Ventures with Acrew Capital, Anthemis, Homebrew Ventures, and Inspired Capital participating. The $25 million Series A followed in November 2021, led by Acrew Capital, with Homebrew, True Ventures, Anthemis, Valor Venture Partners, DCVC, and LowerCarbon Capital joining. Total disclosed venture funding stands at approximately $29.71 million. RLI Corp. made a strategic equity investment of undisclosed size in February 2026 alongside the commercial property product launch; no Series B has been announced.
Products. Kettle's original product was a wildfire reinsurance treaty — cedents (insurers buying reinsurance) cede a portion of their California wildfire exposure to Kettle, which uses its models to price and accept that risk. In 2022 it launched Kettle Re Ltd., a collateralized insurer registered in Bermuda, as a vehicle for transferring a portion of its underwritten risks to capital markets. It later added a parametric hurricane reinsurance product built in partnership with climate data firm Reask, and a California excess wildfire product structured with Amwins and PartnerRe. In February 2026 it entered primary insurance directly with a non-admitted multi-peril commercial property product for California and Nevada, covering wildfire and all other perils, underwritten by Mt. Hawley Insurance Company (an A+-rated RLI Corp. subsidiary) and distributed through wholesale brokers.
Leadership change. In November 2024 co-founders Engler and Manning moved to advisory roles. Isaac Espinoza became CEO. Espinoza spent nearly two decades across Fireman's Fund, Greenlight Reinsurance (where he was a founding member of Greenlight Re Innovations and made early-stage insurtech investments), and Root Inc., where he was SVP of Strategy and Reinsurance and board member of Root Re. He is a Fellow of both the Casualty Actuarial Society and the Society of Actuaries.
Market positioning. Kettle occupies a narrow but real gap: California's non-admitted and reinsurance markets, where admitted carriers have pulled back. Zesty.AI and Cape Analytics sell wildfire-scoring software to insurers. Jupiter Intelligence and Moody's RMS sell catastrophe models. Kettle does neither — it underwrites the risk itself using its own models, which forces it to put capital behind its view. That alignment is uncommon among insurtech analytics vendors. The closest structural analogues are Arbol (parametric climate risk MGA) and Descartes Underwriting (parametric climate MGA, European-led), neither of which operates in the California wildfire reinsurance market.
What it does not do. Kettle does not sell homeowners insurance direct to consumers, does not operate in admitted markets, does not cover non-climate commercial lines, and does not have a US-wide geographic footprint as of May 2026. Its reinsurance product serves cedents — primary insurers — not policyholders directly.
Named deployments
- Mt. Hawley Insurance Company (RLI Corp.) (US)Reinsurance News
- PartnerRe (via Amwins excess wildfire product) (US)Artemis
Known limitations
- Kettle is not a balance-sheet carrier. It operates as an MGA — a managing general agent — and depends on fronting carriers (currently Mt. Hawley / RLI) to hold and transfer risk. That means Kettle's underwriting capacity is subject to those carriers' appetite, which can change if California wildfire losses deteriorate further. (Reinsurance News)
- Geographic focus is narrow. Kettle's core model covers California wildfire and, more recently, Nevada. As of early 2026 it had not expanded to other high-risk wildfire states (Oregon, Washington, Colorado) or to non-wildfire catastrophe perils at scale, despite the founding team's stated ambition to model floods and hurricane wind. (Kettle)
- Total disclosed funding is modest — $4.71M seed (2020) plus $25M Series A (2021) — leaving Kettle relatively capital-light compared with its aspirations to become a broad climate-risk underwriter. The RLI strategic equity investment in 2026 does not have a disclosed amount. No Series B has been publicly announced. (Artemis)
- Kettle's claim that its model predicted 97.23% of California's burned properties over four years is presented on its own homepage without a published methodology or independent audit. External researchers — including Stanford's wildfire science group — confirm that identifying high-risk zones has value, but note that fire ignition prediction at the property level remains probabilistic and difficult to verify out-of-sample. (CNBC)