However, Swiss AI laboratory Jua has recently launched EPT-2, a high-precision AI weather forecast model which the lab claims provides unprecedented accuracy, beating even Big Tech solutions from Microsoft and Google, while using 75 percent less computing power. EPT‑2 delivers both short- and mid-term weather forecasts with unmatched accuracy, providing the intelligence energy operators need to stay ahead of volatility.
REM talked to Marvin Gabler, Jua’s co‑founder and CEO, to obtain insights into how this new generation of physics‑driven AI is helping the energy industry accelerate the clean transition without compromising on stability or profitability.
Can you tell me more about Jua please?
Jua is a Swiss AI lab founded in Zurich that has developed revolutionary weather forecasting technology, creating "Large Physics Models", AI systems capable of truly simulating Earth's physics in real-time rather than just retrofitting AI onto legacy weather systems. Our models currently lead public weather forecasting benchmarks, outperforming AI models from Microsoft and others, as well as public institutions such as NOAA and ECMWF. Jua serves energy traders and grid operators with the world's most advanced weather forecasting AI. Our vision extends beyond traditional forecasting to developing a digital twin of our planet and to become the intelligence layer of decision making of the physical world.
Last June, we raised $11 million in Series A funding co‑led by Ananda Impact Ventures and Future Energy Ventures. In total, Jua has raised $26 million in funding from top VCs like 468 Capital, Promus Ventures, Kadmos Capital, executives from Google Deepmind, Meta AI, Flixfounders, and industry veterans like Siraj Khaliq, one of the founders of the Climate Corporation.
What are the main consequences of climate change for the energy sector?
Climate change is fundamentally reshaping the energy sector by making weather patterns more volatile and harder to predict, creating direct consequences that ripple through every aspect of energy operations from renewable generation planning to grid balancing during peak demand. The growing share of renewables is making the sector more weather-dependent than ever before, as wind and solar generation schedules must be optimised based on accurate forecasts to maximise output and minimise imbalance costs.
Without reliable foresight, grid stability becomes compromised, trading strategies fail to capitalise on market opportunities, and operational planning suffers from uncertainty that can mean millions in lost revenue or unnecessary emissions as backup fossil fuel plants are activated unnecessarily.
What would you say distinguishes a good weather forecast from a bad one?
The difference between a good forecast and a bad one in the energy sector can mean millions in lost revenue or penalty costs in a single day, making forecast quality absolutely critical for modern energy operations. Superior weather forecasting is distinguished by accuracy across both short and mid-term horizons. Update frequency has also become crucial. Our new global weather model EPT-2 updates predictions 24 times daily versus the industry standard of 4 times, allowing energy operators to respond to rapidly changing conditions and optimise generation schedules in real-time.
The best forecasting systems provide uncertainty quantification through probability ranges rather than single predictions, giving energy traders and grid operators the confidence intervals they need for high-stakes decisions about generation dispatch and market positioning.
What are the financial implications of a bad forecast?
Bad forecasts create ongoing operational inefficiencies that compound over time. In renewable energy operations, inaccurate wind and solar forecasts lead to poor generation scheduling, resulting in imbalance costs that can significantly impact profitability, while energy traders face losses from poorly timed market positions based on incorrect weather predictions. Grid operators must maintain expensive backup capacity when forecasts underestimate renewable generation, or risk stability issues when forecasts overestimate available clean energy, leading to either unnecessary costs or emergency measures that can cost millions. To give a concrete example: for a Belgian customer with a small wind portfolio, we were able to reduce balancing costs by 20 percent, leading to a PnL impact in the 7 figure range. For larger portfolios, the impact can be 8 figures and more.
The interconnected nature of energy markets means that forecasting errors cascade through the system, affecting everything from wholesale electricity prices to carbon credit trading, while poor extreme weather predictions can leave infrastructure vulnerable to storms, heatwaves, and other events that cause both immediate damage and long-term operational disruptions.
What kind of growth in the sector are we seeing at the moment?
The weather forecasting sector is experiencing remarkable growth driven by the energy transition and increasing climate volatility, with demand accelerating as renewable energy adoption makes accurate weather prediction critical for grid stability and profitability. The sector is currently undergoing several major transformations: For the first time, due to the AI revolution, private companies like Jua can run global weather models and beat classical numerical weather models built by governments in decades of research. In the past, everyone used the same few weather models, and therefore, if the models were wrong, the entire market made wrong trades. Now, the players adopting AI quick can front run markets or even get tailored models for their specific assets. Also private observation infrastructure is growing rapidly, with the first private weather satellites in space, IoT networks of smart weather stations, as well as long endurance weather balloons.
The global weather forecasting services market was valued at USD 3.2 billion in 2024 and is projected to reach USD 5.8 billion by 2033. The energy sector represents a particularly high-growth segment within this market, as the increasing penetration of weather-dependent renewables creates unprecedented demand for sophisticated forecasting systems that can optimise generation schedules, reduce imbalance costs, and enable proactive trading strategies.
This growth is further accelerated by the need for extreme weather predictions that allow energy operators to safeguard infrastructure and maintain grid stability, with early warnings for storms, heatwaves, and cold snaps becoming essential for operational planning in an era of increasing climate volatility.
Can you tell me more about EPT-2 and also about the main benefits of AI for the sector
EPT-2 represents a fundamental breakthrough in weather forecasting technology, setting a new state of the art in accuracy in both short and mid-term weather forecasts while using 75 percent less computing power than competing solutions from Microsoft and Google. The system functions as a Large Physics Model capable of real-time atmospheric physics simulation, providing the intelligence energy operators need to stay ahead of weather volatility and optimise their operations accordingly.
For renewable energy operators, EPT-2's superior wind and solar radiation forecasts enable optimized generation schedules that increase yield, with customers reporting reduced imbalance costs by up to 20 percent through more accurate production planning.
What’s involved in the development of a digital twin of the planet and what are the main benefits for the sector?
Jua's development of a digital twin of our planet represents the next frontier in Earth simulation technology, essentially the creation of AI systems that can simulate Earth's atmospheric physics in real-time. At the moment, most earth prediction systems are disconnected and ignore that our planet is an interconnected, highly complex system. Global weather models ignore vegetation and infrastructure, wildfire models use weather models as input but mostly ignore atmospheric processes, ocean models are decoupled from higher atmospheric conditions etc., which makes no sense from a first principle but is required because numerical modeling did not allow to simulate the entire planet. This changed with AI. It allows us to build and train models to simulate the entire planet, in its full interconnectedness, which not only leads to better accuracy, but also to simulate millions of scenarios and its impact on weather, resources, wildfires, water streams, vegetation or infrastructure. This ambitious project requires massive data integration from satellites, weather stations, and atmospheric sensors, combined with physics-driven AI algorithms that understand the fundamental principles governing atmospheric behavior.
For the energy sector, this digital twin capability offers transformative benefits, including the ability to predict natural developments and extreme events with highly local precision on a global scale, enabling energy operators to optimise renewable output through superior wind and solar radiation forecasts that account for micro-climate variations. The digital twin approach enables scenario modeling and climate impact assessment that supports long-term strategic planning for the energy transition, helping operators understand how changing weather patterns might affect their operations and investment decisions over time.
Anything else you want to mention?
Jua's achievement of commercial readiness while competitors remain in research phases means energy operators can access this advanced technology today rather than waiting for future developments.
The European sovereign nature of Jua's technology offers energy operators independence from US or Chinese AI models for critical infrastructure applications, which is increasingly important for energy security considerations.
For additional information:
