Red Bull and the Decision Machine: When the Spreadsheet Replaces the Race Strategist at Monza

Verstappen won Monza 2026 by 19.2 seconds. That number is not evidence for AI. It is evidence for a better car at a low-drag circuit, and for a driver who knew how to hand back a position while the rulebook was watching him. But read Red Bull's press release and you'll be told a different story: that a machine told the team what to do, and the team listened.

I have spent a decade reading numbers like these. I do not see an artificial intelligence leading race strategy. I see a rules-retrieval tool repackaged as an enterprise marketing product, and a cloud platform being valued by a lap time rather than a balance sheet.

This is not a story about technology. It is a story about how a racing team packages its own operational credibility to sell to a title sponsor. And if you are a finance professional like me, this is a very well-told sponsorship narrative, anchored to exactly one lap, with no counterfactual, no error rate, and not a single line of data on how often that machine was wrong.

Let's start with the driest point: what this tool actually is.

According to disclosed information, Red Bull's tool began as a search tool helping engineers interpret F1 regulations and past steward rulings. It evolved from document retrieval to in-race recommendation. That is a genuine engineering step, and I won't deny it. But let's name it accurately: a tool that retrieves regulatory text and recommends a course of action is still a regulatory-text tool, with a decision layer bolted on top.

The only demonstrated capability is regulatory adjudication, not strategic optimisation.

Specifically: at Monza 2026, at the first chicane on lap one, Verstappen went off-track while fighting Norris for the lead. The tool recommended handing the position back to avoid a penalty. The team initially disagreed, but ultimately complied. No penalty was issued. Verstappen reclaimed the position and won by 19.2 seconds.

Now analyse that decision as an expected-value problem, not a heroic narrative.

On lap one at Monza, surrendering a position in a heavy-slipstream environment with fifty-plus laps remaining costs a few tenths. A five-second penalty at a circuit where track position is near-decisive costs a two-to-four position swing. The mathematics leans overwhelmingly toward handing the place back. Any competent strategist at any front-running team reaches the same decision — and they reach it on ten years of instinct, not a GPU cluster.

So the real Monza story is not "AI is good." The real story is: a team was willing to follow a recommendation that contradicted its own judgement.

That is the datum worth having. Not the machine, but the structure of human–machine authority on the pit wall.

I encountered this once, at a much smaller scale. In 2026, as an intern at Sanna Khanh Hoa BVN, I audited the books and found the wage bill at 68 percent of revenue, far beyond the 50 percent safety threshold. I proposed cutting star salaries by 20 percent to preserve five billion dong of liquidity. The board delayed, fearing the dressing room. The team finished second from bottom, was relegated, and dissolved with over twenty billion dong of debt.

The lesson I took was not "data is always right." The lesson was: data that is correct but generates no pressure to act is meaningless. Red Bull did the opposite. They had a tool that issued a recommendation, and that recommendation beat human judgement in a real-time, high-pressure situation. From a risk-governance standpoint, that milestone matters more than any winning margin.

But hold on. Hold on.

A team willing to follow a machine does not mean the machine is good. It only means the machine has accumulated enough internal credibility to be granted veto power. Those are two entirely different things, and I have seen no evidence for the first.

At the same time, look at who is telling this story. It is told by Jack Harington, holder of the title "head of technical and performance partnerships" — a commercial role, not an engineering one. He is speaking to CNBC, a business broadcaster, not an engineering journal. And the timing sits in the window of a multi-year title partnership extension with Oracle. The AI story here functions to demonstrate that Oracle's money buys visible, race-winning capability. It is a sales narrative from the sponsorship department, delivered in technical language.

If I were valuing that sponsorship contract, I would note: this is an intangible asset valued by one lap. One lap. No error rate. No latency data. Not a single published case where the machine was wrong or was overridden.

That is not accidental. That is the classic one-sided case selection pattern in every technology narrative sourced from a partnership.

Let's widen the lens to actual race strategy. An F1 strategist solves four problems: tyre-degradation modelling, undercut and overcut windows, Safety Car response, and resource allocation between two drivers within a race. Those four are where strategic value genuinely lives.

Which of those four has Red Bull's tool demonstrated? None. The only evidence is rule adjudication at a lap-one incident.

And here is where I want to push the problem up to a larger strategic question: if the tool were genuinely strong on tyre-degradation modelling and Safety Car handling, why not publish a case in exactly those domains? The silence at precisely the most important points is an information signal, not a random gap.

I have read enough financial reports to know: when a company boasts about revenue but never mentions margin, what it doesn't say is the answer.

Now the most important part — the part the entire article gives one sentence.

The release states that OCI and Oracle AI will underpin "the 2026 debut of Red Bull Ford Powertrains' next-generation hybrid power unit." This is the highest-consequence, least-substantiated claim in the whole story.

Power unit energy management and reliability is the highest-stakes, lowest-tolerance domain in the entire sport. An engine failure on lap thirty loses the whole weekend's points. A single error in the energy deployment map can turn a win into a disqualification. And we are talking about an entirely new power unit manufacturer debuting in the very season the engine regulations reset wholesale, with a historical data pool of limited transferability to the new hybrid architecture.

A marketing release is not technical evidence.

And this is the point I want you to hold onto: a team is using an AI narrative to underwrite confidence in an engine programme that has not yet completed a single racing lap. If that engine has reliability problems in 2026, the AI story will be immediately reread in reverse.

Compare with rivals to sharpen the picture. McLaren partners with Google, described as using "Gemini AI tools" of undefined scope. Mercedes partners with Microsoft, scope unnamed. Ferrari partners with IBM, and this matters, explicitly fan-facing. Three partnerships, three scopes, and only one publicly evidenced as integrating in-race decision-making.

The article folds all four into a single "everyone is doing it" frame. That is a category error. Equating fan-experience AI with in-race strategic decision AI erases a very large capability gap. It is like equating a score-prediction app with a transfer-contract valuation model.

But here I must be fair. There is a genuine asymmetry opening up, and it is not about who has AI.

It is about where the rulebook cannot reach.

F1 has the ATR limiting wind tunnel and CFD runs. It has a token system limiting power unit development. It has a cost cap limiting spending. But there is no mechanism whatsoever limiting the number of strategy-simulation iterations a team can run on a hyperscaler partner's cloud infrastructure. This is an entirely new asymmetry axis, and it tilts toward front-running teams whose title partners are technology giants.

Anyone who has worked in finance recognises this structure. It is a regulatory lag arbitrage. And arbitrage always belongs to whoever spots it first.

I have seen a similar structure at a different scale. Euro 2026, Khanh Hoa had just been re-established, facing a ten billion dong shortfall, and the board wanted to sell the captain in the summer window. I persuaded them to keep the spine of the squad and invest in the academy instead, signing five young players and cutting costs by 20 percent. They survived the season. The lesson there was: in an environment with a hard spending ceiling, advantage comes not from spending more, but from spending where nobody is watching.

Red Bull is doing exactly that. Only at a scale I can only dream of.

So what of the risk? Here I must use the right term: automation bias.

A team set aside its own judgement to follow a model. Under a good outcome, people call that collective intelligence. Under a bad outcome, they call it automation bias. The process is identical; only the result changes. And the published story contains only one side of the result.

The second risk is model brittleness under out-of-distribution conditions. Safety Car, red flag, wet track, a late-race incident. There is no data showing the machine has been validated in those situations. And those are precisely the situations where a wrong decision causes the greatest damage.

The third risk is knowledge concentration. A tool only a handful of people understand from the inside is an operational risk, not an asset. When those people leave, the tool becomes a black box nobody dares switch off.

The fourth risk, and the one few people consider: cloud service valuation under the Financial Regulations. AI and cloud services supplied by a partner must be valued at fair market value. Whether Oracle's contribution sits inside or outside the cost cap, the material does not say. And the precedent from the 2026 cost cap breach shows penalties can include reduced development allowances, not just fines.

This is the highest-probability regulatory friction point in future. Not on-track conduct, but how a technology gift is valued.

On accountability, everything remains reassuringly clear. Under the Sporting Regulations, the responsible entity is the team. A machine-generated recommendation has no independent legal standing. The phrase "it's actually making decisions for us" is commercially accurate, not legally accurate. And that distinction is worth holding onto.

But there is a softer question I want to raise: whether a team running a live penalty-probability model gains an advantage entirely unrelated to car performance. Today that is legal. But it is exactly the kind of advantage that attracts regulatory attention the moment it becomes visible.

And FIA history shows they close such gaps reactively, via Technical Directives, after the grey area has already been exploited.

Back to Verstappen for a moment, because I don't want this piece to slide into dry technology reporting.

What matters about the Monza performance is not the 19.2 seconds. It is that he reclaimed the position on track immediately after surrendering it, and then controlled the race to the point where the final margin became a number beyond argument. A driver placed in a position where he must pay a psychological price, and he processed it like an arithmetic problem.

But I refuse to read 19.2 seconds as a market signal for his contract.

The value of a driver is not in the price, but in how the market revalues him after a major tournament. And a win at a low-drag circuit, in a season where his team's car is superior, is not a market test. It is a performance datum. Reading contract leverage into it is unfounded inference.

I made that mistake once, and I remember it clearly. World Cup 2026, I built an xG model and showed Morocco conceded only one goal before the semi-finals. The most impressive defensive record of the tournament. Yet Hakimi was still valued at around 60 million euros on Transfermarkt. I compiled the data, wrote "Hakimi is worth 80 million euros, not 60," and the piece was widely shared.

But looking back, I recognise something I was not mature enough to see then: I used a four-week tournament to revalue a ten-year career. Exactly as one uses a single lap to value a software machine.

That is why I do not trust the Monza story as evidence of capability.

So if I were an investor or a sponsorship director underwriting the Red Bull–Oracle relationship, I would ask three questions.

First: what proportion of the machine's recommendations did the team override last season. Without that figure, there is no way to value actual dependency.

Second: when the machine is wrong, where and how often. Without failure data, every capability claim is a marketing claim.

Third: how is Oracle's contribution valued under the Financial Regulations. If it sits outside the cap, we are discussing a structural advantage that is unpoliced. If it sits inside, we are discussing an opportunity cost.

Three questions. Not one answer published. And that silence, to me, is the most important datum in the entire story.

But let me be fair once more. There is one possibility I must acknowledge: Red Bull may genuinely be ahead of rivals at the tooling layer, and may be choosing not to disclose detail for competitive advantage. In this industry, concealing capability is a rational strategy. Nobody publishes their tyre-degradation model for rivals to read.

But if you choose to conceal capability, you cannot simultaneously use that capability as a marketing claim. One or the other. You cannot both flaunt and hide. And when someone chooses to flaunt a capability without proving it, what they are selling is not that capability.

They are selling credibility.

And credibility, Red Bull genuinely has. That I must concede. A team willing to follow a recommendation contradicting its own judgement under high pressure is a team that has built a data-driven decision culture rather than an ego-driven one. That is rare, and worth more than any GPU cluster.

But a decision culture is not technology. And this story conflates the two.

Football is where emotion is traded, but professionals must read the balance sheet before reading the scoreline. Here too. You must read the gap between a claim and its evidence before reading a win.

And once you can read that gap, you see the real story is not a race between machines. It is a race between teams to seize the right to shape the story about themselves.

Every record begins with a touch of the ball, and ends with a number on a spreadsheet.

For Red Bull, the number on the 2026 spreadsheet is not 19.2 seconds. It is the sum Oracle paid to attach its name to the moment Verstappen entered the first chicane and decided to hand back the position.

Dissolution is not a full stop; it is the most honest financial report a club ever publishes. So is a claim. A claim without a counterfactual is the financial report of a belief, not of a capability.

And if Red Bull's recommendation list is not even five lines long, it really is just a rough draft presented beautifully. I am not saying it is bad. I am saying a recommendation with no consequences is not a recommendation.

That is why they need a driver clear-headed enough to hand back the position, an engineer capable of calculating that a penalty is less frightening than losing track position.

The Red Bull–Oracle story, to this point, is a sponsorship deal told in technical language. The value is real; it just sits in the intangible-asset line of the commercial department's ledger, not in the strategy model on the pit wall.

And if anyone wants to prove otherwise, they know what to publish. Error rate. Override rate. Performance under chaotic conditions. And one case at a wet circuit.

Until those numbers exist, I will keep this entry filed under "unappraised intangible asset."

And that, to someone who reads ledgers for a living, is an entry worth watching more than any lap.