How AWS is supporting F1’s sustainability drive using the cloud and AI

BlackBook Motorsport talks to Rob Smedley, former Williams, Jordan and Ferrari engineer, and Neil Ashton, principal computational engineering specialist at Amazon Web Services, to find out how the series’ data drive reconciles with its environmental goals.

In recent years, the presence of Amazon Web Services (AWS) in Formula One broadcasts has been unmistakable.

‘F1 Insights powered by AWS’ is now a regular feature on race weekends. With the overarching goal of engaging more fans worldwide, distilling the rather technical nature of Formula One into digestible content makes a lot of sense as part of the series’ growth plan.

The process sees 300 sensors on each car generate more than 1.1 million data points per second, which are then transmitted from the cars to the pit. Unsurprisingly, the data capabilities of the partnership have always been given particular focus.

Whether it’s estimating the number of laps until a certain driver catches up to the car in front or comparing rival car performance in different sectors of a specific circuit, AWS now plays a key role in contextualising the action for viewers. But its partnership with Formula One stretches far beyond the stats.

When the partnership was first announced in 2018, the messaging focused on the ‘never-before-seen metrics’ that would now be on offer. An extension in 2022 saw AWS become a global partner of the series and included a commitment to explore opportunities for sustainable solutions across the sport, building on previous work to reduce freight and personnel travel through remote production capabilities.

With Formula One aiming to be Net Zero Carbon by 2030, this aspect of its relationship with AWS will become increasingly important, especially as it hurtles towards the 2026 engine regulations that are set to introduce 100 per cent renewable fuels to the sport.

But what role will the Amazon subsidiary play in this sustainable future? And how does the firm actively contribute to shifting the environmental narrative around Formula One as a whole?

BlackBook Motorsport sits down with Rob Smedley, former Williams, Jordan and Ferrari engineer, and Neil Ashton, principal computational engineering specialist at Amazon Web Services, to find out.


A post shared by FORMULA 1® (@f1)

How can cloud technology help Formula One achieve its environmental goals?

RS: Formula One visits 23 to 25 countries per year. That’s an awful lot of kit you have to pick up, both from [the series’] point of view and the team’s point of view. You’re going to send [that equipment] all around the world in shipping containers, whether it’s air freight or sea freight.

That includes all the IT equipment. If you imagine the amount of tech and reliance on data that Formula One as an ecosystem has, if you’re doing all that as we were 15 years ago, that’s a massive amount of kit to transport around the world.

[On top of that], when you’ve got back-to-back races, you finish on a Sunday, you’ve got to put it in the back of an airplane on Monday morning, then it flies to the next place to get unloaded on the Tuesday or Wednesday.

As we’ve moved more towards the cloud, that has massive ramifications on sustainability. AWS has massively helped that digital transformation – it’s not only the sustainability aspect, but the latency aspect as well.

We’re not having to send masses of data down a big tube all the way back to [the Formula One headquarters in] Biggin Hill. Now, we’ve got a digital archive all stored in the cloud. All the data we’re bringing off the cars and the timing systems goes into the cloud too.

All of that has a big impact on sustainability: less kit, less people, and you’re generating less heat for that day.

As part of its global partnership with Formula One, AWS was the title sponsor of this year’s Spanish Grand Prix

NA: Formula One and AWS definitely share a vision towards sustainability. I suppose there's two threads: one of those threads has been the work that we have been doing with them on redefining the rules of the sport…to make the racing more exciting.

[Formula One also] could have bought their own large data centre, but independent research has been done that shows running on AWS [records] 80 per cent better sustainability and 80 per cent lower carbon footprint than if they were to do it themselves.

So [the series] reduced the need to do as much physical testing and, on top of that, they could do their simulations in a more sustainable way, because the compute power behind it is from more sustainable sources.

Formula One was also able to move onto our graviton processors, which consume 60 per cent less energy than any of the instances that we offer. These cascading things, which are maybe not as widely known about, actually make a difference.

If you want the sport to be more sustainable, then the tools used to make the sport need to be more sustainable.

Are these technologies helping to reduce emissions in Formula One applicable to the wider automotive industry?

NA: I think the move to digital certification has already begun, so this is particularly relevant in the automotive and aerospace sector, where there's a great desire to move away from flight testing, road testing, physical testing, etc.

And so if you decide that that's true and you want to move to more numerical approaches, you need to find a large scale compute to do it. That is the same for Formula One as it is for the automotive space, you need that compute to be as sustainable, using the lowest energy possible.

They can be running in regions that are using 100 per cent renewable [energy], they can use hardware in instances that we specifically designed to have a low energy usage like the graviton processors. So that means they can then help to achieve that vision of digital certification with green compute to do it.

RS: [When we were trying to design the new car], the simulations [used to take] 40 hours. We then swapped it to the cloud [as it’s] a real high performance compute problem.

Swapping that to the cloud and being able to spin up to 2,000 cores on average, 7,000 cores at peak times – you’re processing billions of nodes of computational fluid dynamics per second, which gets us down from 40 hours to six hours.

That is directly transferable to the OEMs [original equipment manufacturers], although they tend to move at a significantly slower pace as their targets are different to the targets of Formula One.

AWS also has a separate sponsorship deal with the Ferrari Formula One team

How do you see machine learning influencing the future of the sport?

RS: We’re just at the advent of the use and utilisation of artificial intelligence and how that can drive innovations. From a sustainability point of view, it’s about using AI and machine learning to understand the problem.

A lot of businesses try to go straight to a solution. AI is much more important in understanding problems. If you can understand your sustainability issue, if you can define the problem, that allows you to define key solutions that have a massive effect [immediately]. That’s where we’re using AI at the minute.

NA: It’s a really exciting time for machine learning. We’ve seen tremendous growth and interest. We have done work with Formula One, which they spoke about publicly, where we've tried to use machine learning or help them to use machine learning through things like SageMaker, to be able to come up with new designs faster than they could by using traditional simulations.

In three to five years, you'll see an increase in the use of machine learning [in Formula One], but I think it will always be a complement. It will be done to augment, to help to give insights.

– Neil Ashton, Principal Computational Engineering Specialist, AWS

So it's still early days, but I think you see how social media companies are using machine learning to develop better products. I think there's definitely a trend of that also being looked at in the engineering space.

In three to five years, you'll see an increase in the use of machine learning [in Formula One], but I think it will always be a complement. It will be done to augment, to help to give insights. Because of the nature of physics, it's actually very difficult to encode that in a machine learning model.

There’s great potential, but it's more about how can you use machine learning to speed things up, or do things more accurately, rather than a wholesale replacement.