Intrica Computing builds software that makes better trading decisions for energy companies and financial institutions in fast-moving, volatile markets.
To do this, we build tensor network models, a new kind of model at the meeting point of AI and quantum physics. They are calibrated to market data, work through billions of possible futures at once, and compute the best decision with a stated error.
Run batteries? See what they left on the table last year. We compare last year's decisions with ours and with the best possible, on your own data. A few weeks, no change to your operations.
Ask for an auditEnergy companies and financial institutions make these decisions every few minutes. Each one is small, but together they decide whether a battery, a storage site or a trading book makes or loses money.
Estimates for 2030, based on industry forecasts of about €110,000 of revenue per megawatt a year.
Why now. The batteries that will run in 2028 are being financed today, and their owners are choosing software. Power markets moved to 15-minute trading in 2025. And as trading desks adopt AI, they need numbers they can trust.
Machine learning is good at building models from data. Tensor networks, from quantum physics, are good at calculating with such models rigorously, across billions of possibilities. Our tensor network models combine the two: calibrated to market data, then solved rigorously with a stated error.
A model of how prices move, and how uncertain they are, is fitted to market data with modern machine-learning tools.
At their core is the tensor network, a structure from quantum physics that was built to describe systems with more states than could ever be written down.
They compute the optimal decision rigorously and state how far it could be from the true optimum, so it is one you can trust with real money.
The same software runs wherever someone trades on an uncertain future, and we are already testing it in both markets. Each new market needs a new model of that market, not a new engine.
Europe is building grid batteries faster than the software to run them. They are our first market.
Tested on real batteries. In backtests on a full year of real market data, with the day's prices known and net of battery wear, our decisions earned more than a standard planning rule on every large battery we tested. Live trials are next.
Banks need prices and risk figures they can trust, fast enough to act on during the trading day.
Working with a bank. In a first project for a large investment bank, we computed the credit-risk figure it asked for in under a second, on a laptop. Follow-up projects are being prepared.
How clients pay: a yearly licence per asset, with the gain measured on their own data.
A small team at the forefront of tensor-network research, with one of the field's leading researchers as co-founder. Our research at the University of Geneva is funded by Innosuisse, the Swiss innovation agency.
PhD in tensor-network methods, LMU Munich and University of Geneva. Leads the development of Intrica's algorithms.
Research director at CEA Grenoble and director of the Maison du Quantique Alpes. Co-author of widely used tensor-network methods and software.
PhD in tensor networks for quantum simulation. Works on the core algorithms and their applications in energy and finance.
PhD in physics. Fifteen years of computational physics and GPU computing; works on the core software.
Working withUniversity of GenevaHub Quantique @ CEA GrenobleFlorian Stäbler, adviser on power markets
Our missionHelp anyone who trades on an uncertain future make the best decision, and show them how good it is.