IBM and Dallara have proved that a physics‑based AI model can replace hours of traditional CFD simulations with a few seconds of computation. Using a vast dataset from a simulated LMP2 prototype, the new Gauge‑Invariant Spectral Transformer delivers drag and downforce predictions matching conventional CFD accuracy.

The breakthrough comes amid tightening testing budgets in top racing series. Formula 1 and endurance championships cap wind tunnel and CFD hours to curb costs and level competition. Teams now turn to machine learning to squeeze more insight from limited resources, turning a single run into a million data points.

For Dallara, the AI surrogate means faster design iterations and lower computational expense, while maintaining the fidelity needed for high‑stakes racing. Competitors that adopt similar tools can expect quicker validation cycles and sharper aerodynamic tweaks, tightening the performance gap between front‑row and back‑row teams.

The shift to AI‑accelerated aerodynamics signals a broader trend toward data‑driven design across motorsport, where every millisecond counts.