AI Is Rewriting Football: From a 29-Point Offside Line to the Data-Centre Energy Bill

In the 34th minute the ball is in the net. The stand rises, the roar breaks, and then the big screen prints a line of text. Offside. The decision came from 29 data points attached to the players' bodies sampled 50 times per second, plus a sensor inside the ball transmitting 500 times per second. The referee does not run to the touchline to argue. He simply reads out the result the system feeds into his earpiece.

Based on my experience watching matches, this is the fastest I have seen decision-making authority leave human hands in my whole commentating career. Semi-automated offside technology first appeared at a FIFA World Cup in Qatar 2026, then at the FIFA Women's World Cup 2026, in the UEFA Champions League, and from the 2026-25 season in the Premier League. But the part you see — the coloured line on television — is the cheapest part of the story.

The expensive part sits where no camera goes.

Context: a rules fight that does not belong to football

In recent months, according to a Wall Street Journal report relayed by The Independent, a group of technology leaders — Elon Musk (xAI), Mark Zuckerberg (Meta) and Jensen Huang (Nvidia) — met U.S. President Donald Trump to lobby against new AI safety rules. On the other side, Demis Hassabis (Google DeepMind), Dario Amodei (Anthropic) and Sam Altman (OpenAI) warned of risks and proposed a self-regulation mechanism. The whole affair turns on one question: who writes the rules for AI, and how tightly.

Let me be clear: this is technology and policy news, not football news. There is no team, no player, no match in it. But it is the piece the football industry refuses to look at directly, because football has become a major customer of that same industry.

Three layers of dependency have formed over the past decade. The first is officiating: IFAB is the only body that approves technology entering the Laws of the Game, and every VAR system, goal-line technology or semi-automated offside must pass through that door. The second is data: vendors such as Hawk-Eye (owned by Sony), Second Spectrum (owned by Genius Sports), Stats Perform and Sportradar sell tracking data to clubs, broadcasters and betting markets. The third, least discussed, is computing infrastructure: every model needs servers, and servers need electricity.

According to the International Energy Agency's Energy and AI report published in April 2026, global data centres consumed roughly 415 TWh of electricity in 2026, and that figure could reach about 945 TWh by 2030. That is the consumption of a mid-sized industrial nation. Football is not outside that curve.

The European Union brought its AI Act into force on 1 August 2026, with obligations phased in over time. Which means: at the same moment European football is buying more technology, the law governing that technology is tightening.

The three real layers of AI in football

The first layer, on the pitch: what semi-automated offside changes is not the number of goals but the behaviour of defensive lines. When the margin of error is measured in centimetres and measured consistently in every match, the offside trap becomes a far more expensive gamble. A defender must choose: hold a high line to press, accepting a risk of punishment at a precision the human eye can no longer excuse; or drop deep and surrender territory. I have watched several teams choose the second option in knockout ties — not because they feared losing, but because they feared a line.

But here is the core point: technology does not erase controversy; it moves controversy from the referee's eye to the system's calibration contract. When you argue with a referee, you argue with a person. When you argue with a system, you must ask: who calibrates it, how often, where does the training data come from, and who pays for it. No spectator in the stand has access to those answers.

The second layer, in the analysis room: the greatest value of AI in football is not the decision on the pitch, but ownership of the data the match generates. Every match now generates an enormous volume of tracking data: positions, speed, distance covered, pressing density, PPDA — the metric counting the passes an opponent is allowed per defensive action, lower meaning more aggressive pressing. That data sets transfer prices, shapes bookmakers' risk models, and even drives broadcast contracts.

And here is a structural problem I see: the supply of data is concentrated in a handful of vendors. When an entire ecosystem — clubs, broadcasters, betting markets — runs on a single data pipeline, you have a single point of failure. In my trade we call that a single-source story. The report about that White House meeting is the same: every fact flows back to one original report, relayed onwards, with anonymous sourcing. It may be true. But it has no independent verification. Football is now building its operating model on exactly that structure.

The third layer, off the pitch: the energy bill. When the cost of electricity for computation rises, that cost does not vanish — it flows down through contracts. In Europe, some clubs have begun weighing the use of land around stadiums and training centres for server infrastructure. This is my inference, based on how technology corporations lease land near power sources, and I will say upfront: it is unverified. But if it holds, clubs will soon have two new revenue streams — data licensing and infrastructure leasing — and two new revenue streams always come with two new categories of risk.

At this layer I am obliged to talk about the people nobody films. In 2026, I learned to listen to football with my heart when the stadiums had no human sound. From that livestream series I learned one thing: every system on the pitch has somebody operating it. The technician plugging in sensors, the person wiping the surface so the cameras do not drift, the logistics staff running cables. When technology becomes the condition for a match to be valid, those people move from invisible to indispensable — and are still paid the same.

The counterintuitive angle: AI is not smarter, it is just cheaper

There is a very attractive way to tell this story: AI will make football fairer, more accurate, cleaner. I do not believe that telling.

What truly changes football is not AI being smarter than humans. What changes football is AI being cheaper to run. When a system becomes cheap, it does not appear only in the Champions League; it appears in the second tier, the third tier, in youth academies, in places where nobody is capable of auditing the model. That diffusion is not accompanied by matching governance capacity. And when a precise tool is used by an opaque machine, the result is not fairness — the result is a new kind of unfairness wearing a scientific face.

This is also why I do not share the optimism that technology will free the players. A machine measures distance covered, measures acceleration, measures touches, but it does not measure fear. People laughed at me for asking Mbappé about emotion; then the whole of France wept with happiness. I have held the same position since the Neymar affair in 2026: data must be in the piece, but data is never the whole story. Tracking data today is being used to justify congested fixture calendars — while in my long observation it is that congestion, not any failing of the medical staff, that is the single biggest cause of injury.

So where could I be wrong? Two conditions could break this entire argument. First, if the lobbying wave against AI safety rules prevails and the legal framework is delayed or diluted, the technology supply chain in football barely changes, and the compliance cost I worry about never appears. Second, if computing efficiency per watt keeps improving as fast as it has for three years, the energy layer will not be heavy enough to force clubs to recalculate. A woman said one small thing; people laughed. Five years later, they repeated it. I am used to waiting.

What I think will happen, and how to check it

I offer a prediction with a checkpoint: within the next 24 months, at least one of Europe's five major leagues will insert a clause on artificial intelligence and data rights into its collective rights contract. And before the FIFA World Cup 2026 (11 June – 19 July 2026) ends, FIFA will publish a document guiding the use of AI in tournament operations — even if only in the form of principles.

The roadmap I propose has three steps, doable now and needing nobody's permission. One: every tracking-data supply contract must disclose who owns the raw data and for how long. Two: the calibration process for officiating systems must be independently audited and published on a regular cycle, not only after an incident. Three: the employment contracts of the technical operations crew at stadiums must be upgraded in line with how essential their work has become.

Football has never fully decided its own fate. It buys technology elsewhere, rents its laws elsewhere, and pays the price elsewhere. The question of this season is not which team wins the title. The question is: now that the offside line is drawn with 29 data points, who holds the pen — and will the crowd ever be allowed to see that hand?