The Turbo Lag That Came Back: Why Honda's Second ADUO Upgrade Failed to Hold in Singapore

The complaint Fernando Alonso filed after Singapore was not new in its vocabulary: driveability. For Honda, that word collides with a ledger. In Madrid the team spent its second ADUO token of the season on exactly that ailment and fitted a new turbocharger with a single brief — kill the turbo lag. Afterwards both power units went back to Sakura, were opened on the bench, and the finding came back that there was no specific hardware fault. In reporting, that sentence is usually written as relief. To me it is the most uncomfortable sentence in this whole story: nothing was broken, and yet the car was behaving exactly as it had before Madrid.

I went back to the timestamp and found a second story. It does not record itself in any single lap time. Turbo lag is a delay that dissolves into the average of a sector time and shows up plainly in the staircase of throttle application — the driver presses, torque does not arrive, and when it does it arrives in a shove. The peak lap time is the alibi for that delay; the throttle trace is the witness.

Start with the baseline, then ask what changed between Zandvoort and Madrid. Under the 2026 regulations Honda supplies Aston Martin as a works partner, and the entire performance of the AMR26 rests on how cleanly the power unit delivers. ADUO — Additional Development and Upgrade Opportunities — exists precisely for that balance: a manufacturer carrying a performance deficit gets a limited number of upgrade windows, and every token is an entry in a ledger. Honda spent its first token at Zandvoort with a new power unit specification. That specification arrived with turbo lag attached — the spool-up time, the gap between pressing the accelerator and feeling the turbo bite. The second token, in Madrid, brought a new turbocharger built to erase that gap.

Sepang ran better than expected. Singapore was the venue that on paper should have suited the AMR26 — slow corners, traction-heavy exits, a compressive circuit. The result went the other way, and Alonso pointed straight at the power unit afterwards. Aston Martin chief trackside officer Mike Krack later conceded that the Sepang outcome had lifted expectations for Singapore beyond what was realistic. That is the point where I have to separate two things: the pace of the car, and the team's own perception of the car.

What Honda's trackside general manager and chief engineer Shintaro Orihara told Motorsport.com is the key to the whole knot. Even with the new turbocharger fitted, he said, managing the lag in Singapore remained difficult because of the unique track layout and the energy-rich situation there. The consequence was unstable acceleration for the driver. Read that sentence carefully — Orihara is not saying a component failed. He is saying the component did not do its job in the place it was expected to do it. That is a distinction in engineering, and a larger one in reporting.

The mechanism is simple; its limit is not. The MGU-K can be used to compensate for turbo lag — harvesting in the off-throttle sections keeps the engine at a high operation point, the compressor keeps spinning, and the delay on the next throttle opening shrinks. Orihara made clear that this remedy is partial. On energy-rich circuits it becomes difficult to obtain enough MGU-K harvesting value, because the same energy is also demanded for deployment on the straights. When the harvesting budget is stretched, the engine's operating mode is no longer sufficient to control the lag. Turbo lag here is not a component failure; it is a calibration problem produced by an energy-budget shortfall.

That is why the Sakura checks found nothing. On a bench, a power unit meets no energy limit. On a bench there is no street circuit, no kerbs, no walled corridor of nineteen corners. The problem appears when the same unit reaches an environment where the arithmetic of deployment and harvesting is set by the track layout. Krack pointed at exactly that blind spot — simulation does not give a clear picture of how much lag, or the MGU-K cutting out as Orihara described, will cost in driveability. It goes beyond simple sensitivity simulations.

In my experience of the last few years, that blind spot in simulation is the most reliable source of misleading conclusions. A lap-time model treats driveability as noise and folds it into the average; the driver converts the same noise into a cost in confidence. On a circuit like Singapore that cost doubles, because a driver needs to trust the torque on corner exit, and without that trust he opens the throttle late. In a sector time, that lateness looks like a tenth or two. Repeated across nineteen corners, a tenth or two changes a race. I opened a clean spreadsheet and found the crime scene underneath — the offence is not recorded in the lap time, it is recorded in the manner of throttle application.

A model-risk note belongs here. I do not have the full 100Hz trace from this race, and the public sector data is not sufficient for a like-for-like team comparison. So I will not write how much time Alonso lost in Singapore. I will write that Orihara's account and Krack's account of the simulation both point in the same direction, and both are silent in the same place — by how much. As far as my estimate goes, the delta sits in a band I cannot dismiss as noise, and cannot call proven either.

Now the part where I have to be most careful. No hardware fault at Sakura is not good news; it is the most expensive kind of news. A broken part can be replaced, its price is known, and its fix arrives in one delivery. A calibration envelope that fails on certain calendar archetypes is a different thing — it returns at every energy-rich circuit, and every time it returns the team faces the same question Singapore posed. COTA's long straights in Austin are an examination of exactly that kind.

Second, the ADUO ledger. A token was spent, a new turbocharger was fitted, and the problem came back. That is a question of return on investment, and it should be read from the entry, not from sentiment. Honda did not say the turbocharger was wrong; it said its control is not meeting the driver's demands. The next fix is therefore being sought in software and settings rather than hardware. Orihara said plainly that lessons have been taken from Singapore and countermeasures are being prepared, while adding that how those work on the power unit still has to be checked. That sentence is honest and inconvenient at once, because it means the remedy is still pre-validation.

Third, and here I disagree with Krack's framing. He called the Singapore disappointment "emotional dynamics" — Sepang raised the expectation level, it was not delivered, the mood got heavier. That is fair, and from a management perspective it is understandable. But there is a convenient side to turning a technical problem into a psychological one: psychological problems do not appear in homologation paperwork. The car was slow in Singapore because the driver could not trust the throttle, and that trust broke on the energy budget, not on expectation. Managing expectations will help next time; it will not remove the lag.

Correlation and causation deserve a line here too. The good Sepang weekend did not prove the power unit was fixed. One track, one temperature range, one tyre allocation — that is an observation, not a validation sample. I followed one administrative footnote, the ADUO token, and it became a power unit story, because token arithmetic tells you which problem the team believed was real. Honda believed the lag was real. Singapore confirmed it still is.

What to watch in Austin is not lap time. It is the shape of the throttle opening out of slow corners, whether the driver can go to full pedal in one movement, and whether Honda's software countermeasure flattens the staircase of torque delivery or merely shifts where it begins. If the staircase flattens, the problem has been solved. If only the beginning shifts, the team has moved the lag to a new threshold, and the next energy-rich circuit will open it again.

The question is not whether Honda got Singapore wrong. The question is whether a manufacturer that has now spent two tokens can avoid returning to the same kind of circuit with the same complaint — and if it does return, whether the answer will again be that simulation does not show it.