The AI behind Formula One race-day decisions

"When we run our simulations, we run an entire race. We want to run thousands, if not millions of these races. Now, with Azure Batch, we can"

McMillan has been a race strategist for a decade. The softly spoken Englishman was hired by the Mercedes F1 Team after finishing his mathematics PhD at Imperial College London. It was his first job after college and he's been there ever since, moving up to Senior Race Strategist and, eventually, Lead Race Strategist. "We use about five or six strategists a weekend," he says. "Out of those, maybe one or two of them will ever end up being the pit wall strategist who's making the final decision, but you need a whole lot of people running models."

Mathematics is only part of the job. What separates a good strategist from a great one is the ability to function under pressure. "It's not super difficult in the sense that I don't need high-level mathematics, nothing like the mathematics I was doing for my PhD," says McMillan. "What makes a good general strategist? Someone who is a logical thinker, really good at modeling and data analysis. But what really makes you good at the pit job is to be able to make the decision in the moment, to do it very, very quickly, with very good communication skills… when you've got drivers, team principals, race engineers, lots of excitement and intensity. Once you realize you can do that, I think it's pretty clear you're going to be good at this."

Machine learning, as McMillan calls it, has always been a huge part of his job - but in recent years, AI technology has greatly improved his ability to analyze more and more data at faster speeds. "My day-to-day, the actual Sunday doing the strategy, hasn't changed a huge amount yet," he says. So, what has changed? "I write quite a lot of code, but I'm really not a coder. I'm a mathematician, a statistician, a game theorist, but I'm not a coder. I understand what data we've got. I understand what I'd like to do with it. And [with] the advent of AI, like coding tools, I can do those things a huge amount quicker. I can now write models, write analysis, do things at a speed I could never dream of before."

At the start of 2026, Microsoft signed a deal with the Mercedes-AMG PETRONAS F1 Team, a partnership that put Microsoft technology at the heart of its operation. After dominating Formula One for much of the 2010s, winning seven consecutive drivers' championships between 2014 and 2020 and eight consecutive constructors' championships between 2014 and 2021, the team was looking to build on that legacy. With a radically reengineered car performing strongly in testing, there was growing belief that 2026 could see the Mercedes-AMG PETRONAS F1 Team challenge for championships once again.

For the F1 strategy team, one of the biggest benefits of working with Microsoft is access to Azure's vast cloud computing capacity. Using Azure Batch, they can run huge numbers of race simulations simultaneously, testing multiple scenarios before, and even during, a Grand Prix. "Azure Batch has revolutionized our race strategy - scaling thousands of simulations in real time, we can ramp up massive compute during the race so we get… insight when it matters most," says McMillan. "When we run our simulations, we run an entire race. Every lap, every pit stop, every overtake, we've run the entire race beginning to end. We want to run thousands and thousands, if not millions, of these races. Now, with Azure Batch, we can."

In the past, McMillan might run the simulations on a Saturday night while the team was at dinner, which could take up to an hour, sometimes more. Now those same simulations can take seconds and be run from the side of the track. "On a Sunday, in the race, we need those results back as fast as we can," he says.

That data is then fed to the strategy team, who are updating their models live. According to McMillan, despite all the data that is now available to help make decisions, the only time you know for sure if you've made the right choice is at the checkered flag. "By the end of the race, you know the answer," he says. All McMillan and the team can ask for is enough data to take as educated a guess as possible and make a successful final call. And in Miami, they do.