Microsoft has released Aurora 1.5, a major upgrade to its open-source AI weather forecasting model that adds 22 new weather variables, hourly temporal resolution, and ensemble forecasting capabilities. The move marks a significant step toward making advanced Earth-system intelligence available to researchers and organizations worldwide. Yet the release arrives amid troubling news: Microsoft's own greenhouse gas emissions surged 27 percent in fiscal year 2025, driven largely by the massive data center buildout required to power AI applications like Aurora itself.

What Makes Aurora 1.5 Different From Previous Weather AI Models?

Aurora, originally introduced in 2024 and published in Nature in 2025, was already groundbreaking as a foundation model that could be adapted for medium-range weather forecasting, ocean waves, atmospheric chemistry, and climate applications. Aurora 1.5 extends that capability significantly by expanding the model's awareness of atmospheric conditions and adding a feature researchers have been requesting: ensemble forecasting.

Ensemble forecasting works by running multiple simulations with slight variations to show the range and likelihood of possible weather outcomes, rather than relying on a single prediction. This matters enormously for sectors like energy, agriculture, and disaster preparedness, where understanding uncertainty can mean the difference between adequate preparation and catastrophe. The model now tracks:

• Surface and pressure-level variables: Temperature, humidity, precipitation, and radiation fields across different atmospheric layers
• Wind and moisture patterns: Critical for understanding storm development and energy production forecasting
• Hourly temporal resolution: Fine-grained detail that enables precision guidance for events like tropical cyclone landfall or sudden precipitation onset

In testing, Aurora 1.5's probabilistic forecasts outperformed the state-of-the-art European Centre for Medium-Range Weather Forecasts (ECMWF) dynamical ensemble on 88.9 percent of evaluated targets. For tropical cyclones tracked in 2024 and 2025, the model substantially reduced track errors, including roughly one-third lower track error when comparing the ensemble median to the original Aurora.

How Is Aurora 1.5 Being Made Available to Researchers and Organizations?

Unlike proprietary weather models locked behind paywalls, Aurora 1.5 is being released as open source on GitHub with model checkpoints available on Hugging Face, a platform where developers share AI models. This democratization of weather intelligence is intentional. Microsoft is connecting the open research foundation with commercial infrastructure, managed access, and decision-support capabilities through Microsoft Weather services and its Planetary Computer Pro platform.

The strategy reflects a broader philosophy: frontier research should move into practical use. Sridhar Iyer, Corporate Vice President of Microsoft AI, explained the rationale.

"Aurora 1.5 is a meaningful step toward making weather foundation models more open, useful, and practical. By releasing the model openly, we give researchers, developers, and organizations a clearer path to evaluate it, adapt it, and understand where it can help," said Iyer.

Sridhar Iyer, Corporate Vice President, Microsoft AI

The model is already being deployed in real-world applications. The UK Met Office is collaborating with Microsoft to explore how foundation models can work alongside traditional physics-based forecasting systems. Terradot, a company working on carbon dioxide removal through enhanced rock weathering, is using Aurora-derived weather representations to estimate and optimize its climate mitigation efforts under actual field conditions.

Why Is Microsoft's Emissions Spike Overshadowing This Climate Win?

Here lies the paradox: while Aurora 1.5 promises to help organizations make better climate and weather decisions, the infrastructure required to build and run such models is straining Microsoft's own climate commitments. The company reported 21 million metric tons of greenhouse gas emissions in fiscal year 2025, up from approximately 17 million the previous year. That 27 percent increase is largely attributable to Microsoft's rapid data center expansion to support AI workloads.

The breakdown reveals where the emissions are coming from. The largest source was capital goods, which includes server equipment, computer chips, concrete, and steel. Emissions from that category jumped from 6 million metric tons of carbon dioxide to 9 million. About 86 percent of Microsoft's total carbon footprint came from indirect emissions, meaning emissions from things the company buys, sells, and relies on outside its own operations.

Energy consumption in Microsoft's data centers also climbed sharply. The company's electricity use is now more than three times what it was in 2020, reaching 37 million megawatt-hours in fiscal year 2025, up from 30 million the previous year. While Microsoft says it has matched 100 percent of its annual energy consumption with renewable power through purchasing and contracting, that accounting doesn't tell the full story. In Asia and parts of Europe, the Middle East, and Africa, the company couldn't procure enough clean energy locally. As a result, emissions from that category jumped from 259,090 metric tons to 2.7 million.

Steps to Understanding AI's Climate Trade-offs

The tension between AI's potential to solve climate problems and its immediate environmental cost is becoming impossible to ignore. Here's how to think about the competing dynamics:

• Immediate emissions impact: Training and running large AI models requires enormous amounts of electricity. Microsoft's data center buildout, while necessary to deliver services like Aurora, is driving up the company's carbon footprint faster than renewable energy deployment can keep pace.
• Long-term climate benefits: Models like Aurora 1.5 enable better weather prediction, which supports energy grid optimization, agricultural planning, disaster preparedness, and climate mitigation research. These applications could prevent far greater emissions and climate damage over time.
• Geographic inequality: Microsoft can source renewable energy in some regions but not others, creating a patchwork of clean and carbon-intensive operations. This mirrors broader challenges in the global energy transition.

Microsoft Chief Sustainability Officer Melanie Nakagawa acknowledged the challenge in the company's sustainability report, noting that AI is increasing demand for "energy, water, land, and materials," and that "sustainability solutions are not scaling fast enough to meet demand." She declined to say whether Microsoft will miss its 2030 carbon-negative target, instead emphasizing that progress "isn't going to be linear".

What Role Could AI Play in Understanding Urban Climate Risk?

Beyond weather forecasting, AI is also helping researchers understand climate impacts at the neighborhood level. A new study from the University of Illinois Urbana-Champaign, published in Nature Communications, used AI to create the first high-resolution urban air temperature dataset that shows what heat in cities actually feels like to people and infrastructure, rather than what satellites see from space.

The research, led by civil and environmental engineering professor Lei Zhao and graduate student Yiwen Zhang, addresses a critical data gap. International guidelines from the World Meteorological Organization require that standard weather stations be located in open, unobstructed areas far from buildings. As a result, many urban weather stations are at airports or nearby rural land, which don't represent conditions in dense neighborhoods where most people live and work. This gap hinders understanding of public health risks during heat waves, energy demand planning, infrastructure resilience, and climate adaptation.

The study demonstrates how AI can fill observational gaps that traditional meteorology has struggled with for decades. By synthesizing satellite data, sparse ground measurements, and machine learning, researchers can now map urban heat at a granular scale that reflects real human experience.

As climate and weather-related risks continue to affect communities worldwide, the race to deploy AI for climate intelligence is accelerating. Aurora 1.5 and similar tools represent genuine progress in our ability to forecast, understand, and respond to environmental change. But they also expose a fundamental tension: the computational power required to build these solutions is itself a climate burden that tech companies are still learning to manage responsibly.