Short answer: AI weather forecasting uses machine-learning models trained on decades of weather data to predict wind, pressure and other weather. Since 2025, the European Centre for Medium-Range Weather Forecasts (ECMWF) has run an operational AI model — the AIFS — alongside its traditional physics-based model, and research models such as Google DeepMind’s GraphCast and GenCast have matched or beaten traditional forecasts on many measures. For a yacht delivery, AI models are a valuable second opinion on the big picture several days ahead. They do not replace official forecasts, local knowledge or the skipper’s judgement on the day.
Weather decides almost everything on a delivery: when the crew leaves, which route it takes and how comfortable — and safe — the passage is. This guide explains in plain language what AI forecasting is, what it does well, where it is still weak, and how a professional delivery crew uses it.
What is AI weather forecasting?
Traditional forecasts come from numerical weather prediction: supercomputers solve the equations of physics for the atmosphere, step by step. AI models work differently. They are trained on historical “reanalysis” data — a consistent record of past weather — and learn how the atmosphere tends to evolve. Once trained, they produce a global forecast in minutes, using far less computing power. Both approaches still start from the same observations: satellites, weather balloons, ships, buoys and aircraft.
The models sailors hear about
| Model | Developed by | What it is |
|---|---|---|
| AIFS | ECMWF | ECMWF’s Artificial Intelligence Forecasting System, run operationally alongside its physics-based model since 2025 |
| GraphCast | Google DeepMind | A 10-day global AI forecast model, described in the journal Science in 2023 |
| GenCast | Google DeepMind | An AI ensemble model that produces many possible outcomes, published in Nature in 2024 |
| IFS and GFS | ECMWF and NOAA | The established physics-based global models behind most GRIB files sailors download |
What AI models do well
- The big picture: the tracks of low-pressure systems and the position of highs several days ahead — exactly what decides the window for a Biscay crossing or an ocean passage.
- Speed and scenarios: new runs are cheap, so ensembles with many scenarios become practical. They show how certain — or uncertain — a weather window really is.
- Storm tracks: research has shown competitive tropical cyclone track forecasts, relevant for Atlantic crossings in autumn.
Where caution is needed
- Local effects: acceleration around headlands, sea breezes and gap winds such as the Mistral, the Tramontana or the Levante in the Strait of Gibraltar are hard for any global model, AI or not.
- Gusts and squalls: peak gusts and convective squalls are still underestimated by smooth model fields.
- Sea state: waves come from separate wave models — always check them, not just the wind.
- Resolution: many global AI models work on grids of roughly 25–30 km. A GRIB file at that scale cannot show the funnel effect between two islands.
How a delivery crew uses AI forecasts
- Compare models: when AI and physics-based models agree, confidence in a window rises. When they disagree, the skipper plans for the worse case — or waits.
- Read the ensembles: the spread between scenarios tells more than one tidy forecast.
- Check official sources: national met services, NAVTEX and marine warnings have the final word on gales.
- Add local knowledge: pilot books and experience of the area cover what models miss.
- Update at sea: forecasts are refreshed by satellite during the passage, and the plan changes with them.
Read more in our guide to choosing a weather window and our explanation of the Mediterranean’s local winds.
AI beyond the forecast
AI also turns up in routing software that suggests the fastest route through a forecast, in tools that predict fuel use for motor yachts, and in AI assistants that answer sailing questions. They all save time, but they share one rule: a tool can suggest, the skipper decides. A route that looks fast on a screen may cross a tidal race, a traffic separation scheme or an area with orca encounters off Spain and Portugal.
Does AI make deliveries safer?
Better forecasts mean better windows, fewer surprises and less time waiting in port. But the biggest safety factors are still a well-prepared yacht, a well-rested crew and the willingness to wait for the right weather. See our owner checklist before departure and the guide to watchkeeping on a delivery.
Plan your delivery with experienced captains
Our captains combine modern forecast tools — including AI models — with decades at sea. Request a quote and tell us your route and timing.
Related guides
Frequently asked questions
Are AI weather forecasts accurate for sailing?
For the large-scale picture several days ahead, leading AI models perform on par with or better than traditional models on many measures. For local winds, gusts and squalls they share the limits of other global models, so sailors should use them alongside official forecasts.
What is ECMWF’s AIFS?
The Artificial Intelligence Forecasting System of the European Centre for Medium-Range Weather Forecasts — a machine-learning weather model that ECMWF has run operationally alongside its physics-based model since 2025.
Can I get AI weather forecasts as GRIB files?
Some weather apps and routing tools now offer AI model data next to traditional models. Availability varies, so check what your app or routing software supports.
Should a skipper rely on a single AI model?
No. Compare several models, look at ensembles, check official marine warnings and use local knowledge before choosing a weather window.