Seasonal Energy Storage
Batteries bridge hours, but wind and sun vary over weeks and seasons - an affordable TWh-scale store is missing.
Open in the interactive tree →A grid built mostly on sun and wind must carry summer surplus into winter and through multi-day wind lulls. Pumped hydro is big enough but tied to terrain; lithium batteries are too expensive per stored kWh if discharged only a few times a year. Hydrogen in salt caverns is physically possible but loses roughly 60-70% of the energy on the round trip.
As of October 2026
The longest-duration commercial battery technology is Form Energy's iron-air battery (about 100 hours); its first 150-MWh system for Great River Energy in Minnesota is now due online in 2027, the cost target is $20/kWh, and the factory makes 2 GWh per year so far. All grid batteries commissioned worldwide in 2025 add up to 307 GWh, while Germany's gas storage alone holds roughly 250 TWh (about 23 billion cubic metres), around 800 times more. Meanwhile hydrogen, a possible long-term store, is struggling: the IEA cut its 2030 pipeline of announced low-emission hydrogen from 49 to 37 Mt per year.
What is missing
- Capacity costs of a few tens of dollars per kWh or less (Form Energy targets $20/kWh; battery packs are at $70/kWh)
- Better round-trip efficiency: hydrogen reconversion loses 60-70% today
- Hydrogen infrastructure: caverns, pipelines and turbines that burn pure hydrogen
- Market design: who pays for capacity that is only needed a few weeks a year?
- Land and permits for pumped hydro or compressed-air storage on the required scale
Becomes possible once solved
- Electricity supply from 100% sun and wind, even in winter
- Phasing out gas plants as backup
- Energy independence without fuel imports
- Electrifying industry and heating without supply gaps
Open steps
- Ultra-cheap long-duration chemistries Medium AI leverageFind storage chemistries (iron-air, flow, thermal, hydrogen carriers) that cost a few tens of dollars per kWh of capacity and last for decades.
- Hydrogen round-trip efficiency Medium AI leverageRaise the efficiency of electrolysis plus reconversion in fuel cells or turbines, where about 60-70% is lost today.
- Screening underground storage sites Medium AI leverageMap where salt caverns, depleted fields or aquifers can safely hold hydrogen or compressed air at TWh scale.
- Multi-week lull forecasting and sizing Medium AI leverageForecast wind and solar lulls of weeks and size storage and backup for the rare worst cases.
Where AI could help
Medium AI leverage. AI speeds chemistry screening, site search and lull forecasting; cost per kWh, market design and building at TWh scale remain the main hurdles.
- Screen electrolytes, redox molecules and materials for cheap long-duration chemistries
- Run self-driving labs that test hundreds of formulations per week
- Size and dispatch storage in energy-system models to find the cheapest mix
- Forecast multi-week wind and solar lulls to plan reserves
Shown so far
- In 2026 an Argonne robotic lab with machine-learning-guided selection ran 6,000+ experiments in five months on organic flow-battery solvents and found a common degradation pathway limiting lifetime. source
- In January 2024 Microsoft and PNNL screened 32 million candidate materials down to 23 in about 80 hours and built a working solid-electrolyte battery prototype with less lithium (battery materials, not seasonal storage). source