Microsoft opens up its Aurora weather foundation model

Aurora is Microsoft’s foundation model for Earth systems and atmospheric science, designed to handle global-scale weather prediction, hurricane tracking, air quality forecasting, and ocean wave modeling.

What it does and how it works

Aurora operates as a massive foundational backbone rather than a single-purpose forecasting tool. It was trained on one of the largest collections of atmospheric data ever assembled, learning the complex physics and continuous flows of planetary systems.

Because the baseline model captures general atmospheric dynamics, it adapts easily to downstream specialized tasks. Researchers can take the pre-trained checkpoint and fine-tune it with relatively modest amounts of custom data to predict specific phenomena, like local wave dynamics or urban air pollution levels.

Microsoft is also tying this work to SPARROW, an initiative exploring low-cost, community-deployable open-source weather stations. The goal is to plug real-world observation blind spots in underserved regions so models like Aurora have reliable ground data to work with.

Availability and what is next

Aurora's source code and model weights are already public. Microsoft is building out the next phase alongside Professor Rich Turner’s machine learning lab at the University of Cambridge.

Moving forward, Microsoft pledged to keep future model releases fully open-source alongside all new derivative architectures. More importantly, the team will open up the entire end-to-end training pipeline, allowing external teams to replicate and adapt the training mechanics rather than just downloading frozen weights.

The immediate focus is empowering national meteorological agencies in both the Global South and Global North. Instead of spending millions building custom weather infrastructure, regional teams can take the Aurora base and calibrate it to their local microclimates.

Researchers and engineers can access the existing open-source repository today, with new pipelines slated to roll out directly into the public project.

Why it matters

If you work in environmental data, renewable energy planning, or climate risk modeling, Aurora is worth serious attention. Energy traders and grid operators are already eyeing its forecasting capabilities to better balance renewable generation and protect infrastructure ahead of severe weather events.

While competing tech giants often guard their advanced Earth-system models behind closed research papers or proprietary APIs, Microsoft is giving away both the checkpoints and the training pipeline. That completely lowers the entry barrier for scientists who need enterprise-grade forecasting without enterprise-scale budgets.