Oscar Piastri lost two tenths to Lando Norris in qualifying at Spa-Francorchamps, and McLaren team principal Andrea Stella says it had nothing to do with how Piastri drove. The culprit is the 2026 power unit's machine learning software — and it is confounding engineers and drivers across the entire field.
Stella was unambiguous. Comparing Norris and Piastri's best Q3 laps, the Australian was losing time on the final straight and through Blanchimont because of how the power unit operated, not because of anything he did with his right foot. "Oscar is losing time in the final straight and Blanchimont for reasons that have nothing to do with Oscar's driving. They are just a minor deviation in how the power unit was operated," Stella said. He added that Piastri could have been one to two tenths faster had the power unit behaved as anticipated.
The problem is structural. The 2026 hybrid units do not deploy energy in a fixed, predictable pattern. They calculate deployment in real time, building an information bank corner to corner and predicting future energy needs on the fly. Any disruption to that learning loop — a missed braking reference, a run through gravel, lost track time — can cascade across multiple laps and even multiple runs. Piastri suffered a hydraulic leak on Friday and went wide on his first Q3 attempt. Both incidents likely fed bad data into the system.
Stella confirmed the same pattern is visible at Mercedes. Kimi Antonelli, who leads the championship standings on 179 points, and team-mate George Russell — separated by 25 points at the top of the drivers' table — show identical straight-line deviations when their data is overlaid. Russell confirmed qualifying preparation had been dominated entirely by solving the straight-line speed mystery, with set-up and tyres secondary concerns.
