Eight AI partnerships signed in six months. Williams runs Claude. McLaren runs Gemini. Red Bull runs Oracle. The 2026 regulation overhaul has turned the paddock into one of the largest live commercial AI deployments in sport.

The teams in the Formula One paddock have always quietly run on data. They have just become noisier about it. According to a Reuters report wired through Sunday, eight new AI partnerships have been signed across F1 and its 11 teams in the last six months alone, and the technology category, broadly defined, has overtaken almost every other line in team budgets.

AI and machine-learning brands now account for four of the top 15 new sponsorship investors in the sport. Among the headliners: Williams running Anthropic’s Claude, Red Bull deepening an Oracle relationship that has shifted from search-style queries into agentic decisioning, and McLaren’s long-running Google partnership migrating from Pixel hardware to Gemini.

It is, depending on how one frames it, either an inevitable convergence or the moment a sport that has always been an engineering exercise became, openly, a software exercise too.

The numbers behind the noise

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

The data Reuters cites is from research firm Ampere Analysis, and the broader spending picture is striking. Technology now leads the top 10 spending categories for F1 teams, reaching an estimated $769m last season, up 41 per cent on the year before. Yahoo Sports’ carry of the Reuters wire breaks out the partnership table in detail, and the cluster of AI-and-cloud names is now meaningfully concentrated. CoreWeave, the GPU-cloud operator, has joined Aston Martin.

Oracle has expanded inside Red Bull. Anthropic, a relative newcomer to motorsport sponsorship, has placed Claude inside Williams’ operations and race-strategy stack. Google has rolled Gemini into McLaren’s analytical platform. Each of these is, in its own way, a marketing bet for the brand involved. Each is also a serious operational commitment.

That dual function, sponsor logo and operational software, is the defining feature of what has changed. The historic F1 partnership template was a logo on a sidepod and a hospitality suite. The 2026 template is closer to a deployed enterprise contract. Anthropic engineers reportedly sit with Williams’ strategy team.

CoreWeave compute powers Aston Martin’s CFD pipeline. Oracle’s agentic systems shape Red Bull’s pit-wall decisions. The sponsorship is the production deployment.

The proximate driver is the 2026 technical regulation overhaul, the largest rules reset F1 has experienced in over a decade. The new chassis-and-power-unit specification has changed the maths of car development in ways that favour teams able to evaluate thousands of design variants quickly, and disadvantage those reliant on physical wind-tunnel hours.

Racing Bulls partnered with Neural Concept ahead of the regulation change specifically to use digital twins and machine learning to evaluate aerodynamic configurations that would be impossible to test physically inside the FIA’s restricted testing windows. Most teams have done some version of the same.

The wider point is that F1 has, since the 2022 budget cap, become a sport in which competitive advantage is constrained by money, computation, and access to talent in roughly that order.

IMD’s recent analysis of F1’s human-AI edge framed the dynamic as an industry case study: under cost-capped conditions, the teams that win are the ones that extract the most decision quality per dollar of compute, not those that simply spend more. Generative AI, deployed inside strategy rooms, race-engineering desks, and CFD pipelines, fits that constraint exactly.

What the AI is actually doing

On race weekends, the visible work is real-time. McLaren, by Google Cloud’s own description of the team’s setup at Miami, runs close to 300 million race simulations before a Grand Prix begins, with generative models surfacing pit-window options and tyre-compound combinations that would be impossible for a human strategist to evaluate within a single weekend’s time budget.

McLaren’s chief AI officer has described the accuracy of the resulting predictions as approaching “an almost eerie level” alignment with what the actual race produces.

Behind the visible work is a less glamorous one. Formula 1 itself, the rights-holder rather than any individual team, has built generative AI workflows on AWS to accelerate race-day issue resolution: triaging telemetry anomalies, surfacing broadcast-relevant context, and reducing the latency between a pit-lane incident and a televised explanation of it.

Lenovo’s ThinkPad X9 Aura Edition was trialled at the Chinese Grand Prix earlier this year, running a MATLAB sports-data model that produced production-floor insights more than 30 per cent faster than a non-AI laptop, the kind of incremental gain that, repeated across 24 race weekends, adds up.

And then there is the regulator. Motorsport.com has reported that the FIA itself is deploying AI to police one of the sport’s most contentious technical questions in 2026, with details closely held but the broad direction visible: rule enforcement, like rule design, is becoming algorithmic.

Formula1.com’s profile of the FIA’s new “tech director” AI expert framed the appointment as part of the same shift, intended to give the regulator the analytical capability the teams already have.

Williams, Claude, and the consulting-arm logic

The Anthropic-Williams partnership, in particular, fits a wider strategic pattern at the model company. TNW reported earlier this week that Anthropic is launching a $1.5bn enterprise AI services firm with Blackstone, Hellman & Friedman, and Goldman Sachs, designed to embed Claude inside the operating businesses of major buyout firms.

The Williams deployment is, in operational terms, a pre-cursor: a high-profile, demanding customer using Claude inside a real-time decision environment, in front of cameras. Whatever the team learns about Claude’s performance under race conditions, Anthropic’s enterprise customers will hear about second.

That is a feature of F1 sponsorship that AI vendors have realised faster than most: the sport is, by structure, a public technology bench. Cars race for two hours every other weekend, and what works on the pit wall is, eventually, written up by every business publication that covers the team.

For an enterprise software vendor trying to demonstrate that its tools work in adversarial, high-throughput environments, F1 is one of the best demos available.

The data the cars themselves now produce

The underlying dataset has grown to a scale that makes the AI investment look proportionate rather than indulgent. TNW reported the McLaren data scale several years ago, when each car was generating roughly 250 million data points per race. Current estimates, across newer sensor packages, put the figure higher.

Cars in 2026 carry between 300 and 600 onboard sensors and stream more than a million data points per second; aerodynamic, mechanical, electrical, thermal, and driver-input channels all feed into the same telemetry pipe.

Mercedes uses G42’s predictive algorithms layered with SAP enterprise systems. Ferrari has built customised models on Amazon SageMaker, achieving up to 60 per cent faster CFD simulations for component testing. McLaren runs Dell’s portable micro-datacentres trackside to update the car’s digital twin in real time. The setups differ. The category does not.

Even TNW’s earlier piece on AWS’s analysis of the fastest F1 driver of the past 40 years hinted at where this trajectory was heading. That story used machine learning to extract a single comparative metric across 40 years of qualifying data. The 2026 version is the same approach applied at every level of the operation, in real time, by every team in the paddock.

What it does not solve

There are limits worth naming. The first is the talent problem. F1 race strategists, like fighter-jet pilots, develop intuitions over years that no amount of compute fully replicates. TNW has previously written about the limits of pure-AI competition in the context of the Indy Autonomous Challenge, where the absence of human dramatic stakes turned out to matter more than the engineering achievement. F1 has not, by anyone’s account, removed humans from the pit wall. It has armed them.

The second is the cost-cap question. F1’s $135m team-cost cap was designed, in part, to constrain the kind of compute-and-engineering arms race that AI deployment now arguably revives.

The accounting treatment of partner-supplied AI infrastructure, donated GPUs, embedded engineers, in-kind cloud credits, has become an active topic in the paddock. Whether the FIA tightens those rules, or accepts that AI sponsorship is the next frontier of legal cost-cap workaround, is one of the open governance questions of the season.

The third is competitive convergence. If every team has access to roughly the same generative AI capability, the marginal advantage from deploying it shrinks over time.

The teams that win in this environment will be those that build distinctive proprietary models on top of public foundations, the way Ferrari has invested in correlation-fixing models that close the simulator-to-track gap, or that find ways to deploy the technology faster operationally rather than more elaborately.

Where this points

Three signals will indicate whether AI in F1 is a lasting competitive lever or a fashionable line item. The first is whether the teams currently behind on the grid, financially or technically, can use AI partnerships to close gaps that money alone could not previously buy.

The second is whether the regulator’s own AI capability matures fast enough to police the technical and financial questions the teams are now asking it. The third is whether the broadcast experience, where AI-augmented insights and faster contextualisation of incidents are visibly changing what fans see on screen, develops the same competitive importance off-track that telemetry models already have on it.

On Sunday in Miami, Kimi Antonelli won. By the time the race ended, Williams’ Claude-supported strategists had run several million scenarios against the actual unfolding telemetry; McLaren’s Gemini-driven simulations had been tested against three of their own pit decisions; Red Bull’s Oracle stack had recommended several agentic interventions on a car that, by the chequered flag, finished where the model expected.

The race, in the conventional sense, was won by a 19-year-old in a fast car. The race, in the broader sense, was won and lost in front of screens by people the cameras almost never show.

F1, like the wider industry it now mirrors, is in the middle of figuring out which of those races matters more.