Foresight Grid built a specialized AI foundation model that learns the physics of the electricity grid to forecast wholesale prices and optimize trading decisions for generators, traders, and large energy consumers. We are focused on Texas, the most volatile electricity market in the US, where over $100 billion in energy is traded annually across physical and derivative markets.
Across CAISO, ERCOT, and PJM, wholesale electricity is a $100B+ market, with $370-530B in derivatives layered on top. Yet every grid participant makes decisions using tools designed for a dispatchable grid that no longer exists. The result: billions in missed opportunities, mismatched hedges, and mispriced risk.
"Forecast differences between day-ahead and real-time markets have grown so significant that RTOs across the country are being forced to create entirely new ancillary service products just to manage the uncertainty created by variable energy resources."
Source: FERC, 2023 State of the Markets Report, pp. 10-12 (March 2024)
https://www.ferc.gov/reports-analysis
ERCOT Day-Ahead to Real-Time spreads can swing from -$4,000 to +$28,000/MWh in a single day. While prices themselves are capped, the DART spread is not. Legacy forecasting tools predict averages. Markets reward those who predict the spreads.
ERCOT runs 288 five-minute dispatch cycles daily. Battery operators must decide when to charge, discharge, or hold, often without knowing what the next hour looks like. The winners have better foresight.
ERCOT restructured its market on December 5, 2025. Statistical models may need 6+ months of new data to recalibrate. Physics-based models adapt in days. That gap is where money is made or lost.
Forecast errors don't just cost traders money. They inflate the risk premium on every solar project, every storage asset, and every renewable investment decision made on the grid.
Foresight Grid is building a purpose-built AI system designed from the ground up for wholesale electricity markets. Unlike general-purpose AI tools or legacy statistical methods, our model is trained directly on grid physics, market structure, and real-time settlement outcomes.
We own and operate every layer of the system: the model, the retraining pipeline, and the inference deployment. No third-party model dependency. No black-box APIs. Our customers get full transparency. We maintain complete control over performance.
Our initial focus is ERCOT, the Texas wholesale market. It is one of the most dynamic, fastest-growing electricity markets in the world, and the most volatile in the United States. From there, we plan to expand across US deregulated markets and eventually into volatile wholesale markets globally.
Foresight Grid serves power plant operators, including wind and solar, battery storage owners, virtual traders, and Qualified Scheduling Entities who need AI-grade decision intelligence to compete in volatile wholesale markets.
33 years in the solar industry. Miguel has worked for SunPower, First Solar, and Solar World, and led projects across 32 countries. He previously founded 8760 Consulting (acquired) and brings deep expertise in energy project development, M&A, and commercial strategy.
PhD in Electrical Power Engineering with doctoral research focused on AI-based solar power prediction. Published in Nature Scientific Reports, IEEE, and AIP Publishing. Her work sits at the intersection of deep learning and power systems, a rare combination that is central to our technical approach.
Supported By
We are working with solar, storage, and virtual trading operators across ERCOT. If you're curious about how AI-native forecasting could work for your assets, we'd like to hear from you.
Contact Us