About
Silurian builds foundation models that simulate Earth, beginning with global weather forecasting. It sells access to these models through an API for customers in sectors including energy, insurance, agriculture, and logistics; its differentiation is high-resolution, long-range Earth simulation intended to improve decisions beyond traditional numerical weather prediction.
Market
Silurian competes in AI-powered weather intelligence, weather forecasting APIs, and physical climate-risk analytics. Its positioning is enterprise- and asset-focused: GFT combines atmospheric and land-surface data with SCADA feeds, outage logs, sensor telemetry, and other infrastructure layers, enabling site-specific forecasts rather than generalized regional predictions. The company differentiates through customizable foundation models, rapid-refresh and high-resolution forecasts, and direct integration into operational use cases such as renewable-energy scheduling, energy trading, and grid reliability.
Silurian primarily targets enterprise asset and infrastructure operators, especially energy providers, utilities, renewable-energy schedulers and traders, and organizations involved in disaster response. Likely buyers include grid, weather-risk, operations, and energy-management teams that need hyperlocal forecasts tied to infrastructure data and decisions.
At a Glance
Problem
Silurian addresses the gap between weather information and decisions that depend on it. Insurance companies must estimate hurricane risk, airports struggle to anticipate delays, and utilities face potentially costly outages when conventional forecasts are too coarse, too slow, or insufficiently localized. The underlying economics are unfavorable: traditional numerical forecasting can take one to three hours for a global 9-kilometer forecast, while increasing resolution can multiply compute and energy costs by roughly eight times. A particularly compelling use case is electric-grid reliability, where missed forecasts can cause asset damage, service disruptions, blackouts, and higher dispatch costs.
The company’s core insight is that enterprises do not primarily need another generic weather map; they need forecasts of how weather will affect their specific assets. In the Hydro-Québec collaboration, for example, Silurian focused on predicting icing on transmission conductors and other grid variables far enough ahead for operators to act, rather than merely reporting regional freezing conditions.
Product / Service
Silurian builds and customizes Earth-system foundation models, starting with weather. Its Generative Forecasting Transformer, or GFT, is a 1.5-billion-parameter model that forecasts global conditions up to 14 days ahead at approximately 11-kilometer resolution. The company delivers the model through its Earth API, which provides global, hourly forecasts and variables relevant to renewable energy, including 100-meter winds and surface solar radiation; its public product also offers a playground and software-development kits.
The differentiator is customization on customer data. Silurian can post-train GFT on sensor feeds, historical outcomes, SCADA data, outage records, and other infrastructure information to produce hyper-local, asset-aware forecasts and downstream impact variables. In Hydro-Québec testing, the customized model reduced error by 15% for surface temperature, 20% for hub-height wind, and 35% for precipitation versus a legacy baseline, while achieving 0.72 average precision for rime-ice detection and actionable warnings up to 72 hours ahead. This is intended to turn weather forecasts into operational recommendations while lowering latency and compute costs.
Market
Silurian competes in AI weather forecasting, Earth-system foundation models, and enterprise weather or climate-risk intelligence. Its target verticals include energy and grid operations, insurance, agriculture, logistics, aviation, and government. The company positions GFT against both established numerical systems—such as ECMWF’s HRES, NOAA’s HRRR, and Europe’s ICON—and other AI forecasting models including Google DeepMind’s GraphCast; the broader AI-model set includes Aurora, FourCastNet, and Pangu-Weather. Commercial weather-intelligence platforms such as Tomorrow.io and The Weather Company are adjacent alternatives for business users.
Silurian is not merely pre-product: YC lists it as an active company founded in 2024, its GFT API has been made publicly available, and it has announced a Hydro-Québec grid partnership and selection for TotalEnergies’ electricity and renewables accelerator. Its public materials show validation and deployments or programs in energy, plus evidence across energy and government, but do not disclose revenue, customer count, or a confirmed paid-commercial scale; therefore its commercial traction is best characterized as early validation and partnerships rather than established scale revenue.
Founders & Leadership
Funding History
Y Combinator, Pioneer Fund
Recent News
NOAA’s Atlantic Oceanographic and Meteorological Laboratory and Silurian AI established a CRADA to advance AI-driven tropical cyclone forecasting. The work will create an ML-ready cyclone database and refine Silurian’s Generative Forecast Transformer for improved storm-track and intensity prediction.
Kirkland-based Silurian raised $6 million to develop AI foundation models for weather forecasting and Earth-system simulation. Its Generative Forecasting Transformer integrates atmospheric, land-surface, and infrastructure data for applications including energy, insurance, logistics, and climate technology.
Silurian announced a GFT-HQ capability that learns rime-ice risk natively and provides operators with full-day warnings aligned with utility planning windows. The post presents the technology as a way to move transmission-grid operations from reactive monitoring to proactive intervention.
Silurian described its collaboration with Hydro-Québec to post-train the Generative Forecasting Transformer on utility-grade observations and forecast asset-level grid risks. The resulting system covers variables such as temperature, precipitation, wind, turbine icing, and conductor rime ice, with reported improvements over numerical weather prediction models.
Silurian showcased GFT-C’s performance on contrasting 2025 tropical cyclones: the model predicted Tropical Storm Henriette’s dissipation two days ahead of the official forecast while also tracking Hurricane Erin’s rapid intensification. The post uses the storms to demonstrate GFT-C’s predictive capabilities across ocean basins.
The Center for Data Innovation interviewed Silurian CEO Jayesh Gupta about the company’s Generative Forecasting Transformer and its use of hyper-local and operational data. Gupta discussed applications in energy and emergency management, including transmission-line icing forecasts and the company’s longer-term decision-optimization vision.
Active Roles
1Business Model
Silurian appears to monetize customer access to its weather and Earth-simulation models through its Artificial Planetary Intelligence API, which lets customers query models for locations around the world. Detailed pricing—such as subscription, usage-based, or enterprise fees—is not disclosed in the available evidence.