The Mercedes-AMG Petronas Formula 1 Team is using HPE high performance computing technology to test car designs faster, cutting pre-race simulation time from three days to one.
To date, the team’s two drivers have won 12 Grand Prix this season, including Kimi Antonelli topping the podium at the Circuit of The Americas (CoTA) in Austin, Texas on 25 October. It was the ninth win of the season for the 20-year-old Antonelli, with teammate George Russell taking the other three.
At an event held before the race in Austin, Ruth Buscombe, a UK motorsport engineer, strategist and television broadcaster, told analysts and journalists this year’s rule changes had been the biggest since 2014.
Among the new rules, the cars’ power units are now split evenly between internal combustion and electrical power, and they must run on 100% sustainable fuel. The chassis is lighter and smaller, and the cars no longer generate most of their downforce through the floor.
All 11 teams also operate under the same financial cost cap, designed to limit spending and create a level playing field.
Those changes led to uncertainty for each team at the start of the season. Toto Wolff, the Mercedes-AMG Petronas team’s principal and CEO, describes a complete rethink of how the cars are operated, while a moderator at the event noted there is no in-season testing to fall back on.
Mercedes-AMG Petronas, though, got off to a strong start. Wolff stated the team “got it right from the get-go”.
Michael Taylor, the team’s IT director (pictured, left), said every aerodynamic idea the team develops starts with high-performance computing, which is where HPE’s work comes in.
HPE said in a March 2026 update its new computing environment cut pre-event simulation time from three days to one.
Wolff calls HPE “an integral part of our IT and technology story.” He is also clear data doesn’t make the decisions. He recalls an upgrade this season which was expected to cut three-and-a-half tenths of a second from a lap but left the car three-and-a-half tenths slower.
The sensors showed the aerodynamic load was as expected, and the engineers said everything was as they had thought. But the driver said the car was unpredictable and he had no confidence in it.
When data and driver disagree, Wolff’s default is to trust “the most expensive sensor,” the driver sitting in the car.
Taylor said the job on the IT side is to make sure data can be trusted in the first place, with reliability ahead of performance.
How it works
HPE provides the compute, networking, hybrid cloud and security infrastructure behind the team’s engineering and race operations.
On the simulation side, AMD-based HPE ProLiant servers and HPE GreenLake replaced the team’s previous high-performance computing infrastructure, and computational fluid dynamics (CFD) work that shapes the car’s aerodynamics runs on them.
Max Walker, a chief technologist at HPE (pictured, left), has been part of the work with the F1 team from the start.
His remit covers HPE’s most strategic customers across the UK, Republic of Ireland, the Middle East and Africa.
“I kind of focus on the holy trinity of IT, so people, the process and the technology,” he said.
Walker describes the work as finding marginal gains and stated HPE’s contribution is “how we add our tenth-of-a-second to the performance of the team”.
Walker cited engineering as a common focus for HPE and Mercedes-AMG Petronas.
Testing, simulation
Walker said the new CFD platform replaced a legacy cluster running OpenFOAM, the open-source software the team uses for its simulations. HPE ships devices to the team, which benchmarks them in its own test harness before a full implementation is built.
“If that performance profile shows a benefit, a percentage increase, then we look to build a solution around it,” Walker said.
For Mercedes-AMG Petronas, the testing window is the gap between races.
“They could run more simulations in that time frame,” he said.
Walker explained “the CFD platform is like a digital twin of the car, or a component they want to test,” such as a front wing, brake duct or suspension part. It assesses changes in airflow and their potential impact on performance.
If the numbers look good, the part is made in carbon fibre, milled titanium or 3D-printed metal. It then goes through the wind tunnel and finite element stress testing before it reaches the track.
“That is a continuous loop of development throughout the whole season, 24-7. It never stops,” Walker said. “The car that starts at the start of the season isn’t the same car at the end of the season.”
Data
Walker said an F1 car typically carries around 200 sensors, which is roughly 18,000 channels of data. With the extra aerodynamic rakes teams run in practice to check airflow over the tires, that rises to about 400 sensors and 36,000 channels.
In the garage, “we have an umbilical cord, an Ethernet cable, essentially, that plugs into the car,” Walker said. It pulls between 2 and 5 gigabytes off the onboard computers over Aruba CX switches. Around 100 of the team’s race engineers at the circuit access the data wirelessly.
From there it travels over SD-WAN, compressed and optimised, to the team’s Brackley, UK, headquarters. It is also mirrored to the powertrain site at Brixworth, so engine and gearbox engineers see the same numbers in real time. The wide-area links come from a third-party provider, Walker said, because race circuits supply only part of the infrastructure.
Walker offers an example from Brazil, a few years ago. A warning from an engine sensor appeared on the steering wheel while Lewis Hamilton was driving for Mercedes-AMG Petronas. The pit wall saw it immediately, and the data went back to the factories. Walker recalled the sensor pointing to a serious fluid leak, the kind of fault which could have ended the race.
“In four corners, the data had got back, they spotted it in turn three,” he said. “By turn seven, they changed the engine mode, based on that data, to reduce the chance that the engine would destroy itself.”
History
When the partnership began, HPE’s assessment found a lot of ordinary PCs running at the circuit.
“Albeit that it worked, I think we had a much more eloquent and ruggedised platform, a more performance-dense platform,” Walker said.
The team moved to HPE Edgeline and added an Aruba Wi-Fi network, which meant greater available workload at the track and higher reliability.
“Consistency is key to unlocking performance,” he said, adding it let the team analyse data faster than its competitors.
At the factory, the aim was an infrastructure which largely runs itself. HPE built a software-defined data centre on Aruba CX networking and HPE Synergy compute, which Walker calls “the Swiss Army knife of platforms” because it can run bare metal, virtualisation or containers. An open-source automation layer on top lets the environment spin-up to meet engineers’ needs without manual work.
Self-driving networks, security
HPE is investing heavily in operational AI, Walker said, with agentic tools which watch network performance and recommend fixes, for instance flagging a saturated link and suggesting a quality-of-service change, with a person still in the loop.
“We don’t have to have humans’ eyes on glass anymore,” he said. “We have agentic bots that are doing the eyes on glass, but then prompting the human to do something about it.”
On security, Walker noted the team was once customers of Silver Peak, an SD-WAN specialist HPE acquired, and that the technology now sits within the vendor’s Aruba SASE portfolio.
HPE stated network access controls, encrypted connections and cloud-delivered security help protect distributed operations like the team’s.
How ideas get funded
HPE and Mercedes-AMG Petronas meet regularly in London for an innovation forum. They discuss upcoming HPE technology and the team’s requirements. Proposals need a business case showing a measurable benefit in operational efficiency, performance or sustainability, and must fit within the sport’s cost cap.
Walker compares the environment to a bank or a life sciences company, heavily regulated with finite money and people. If a new wing or brake duct can shave a tenth or a hundredth off a lap it will win investment over anything in IT, he said, unless IT can make its own case as it did with the CFD platform.
Those decisions are made quickly by Wolff and his management team, Walker said. “They’re very agile.”
The faster simulation platform came out of those discussions, Walker said. So did plans to use HPE Morpheus Enterprise, which he said will let engineers request infrastructure from a catalogue through automated workflows without adding to the team’s operational workload.
It also supports lower-cost virtualisation, “meaning that the team can do more for less,” Walker said.
Roadmap
Walker said Mercedes-AMG Petronas would probably be the first customer in the UK to receive a Gen 13 server, which the team will use to test performance, sustainability and workload density.
“We are seeing massive performance gains for Gen 13, and we see it’s got 50% more memory bandwidth,” he said. “And in a CFD workload, that’s a big deal.”
He noted the team has also worked with HPE on machine learning and reinforcement learning, with an emphasis on “consistency, reliability, and performance”. For some simulations, the aim is now to bring turnaround below a day.
The post Feature: How HPE fuels Mercedes-AMG F1 engineering appeared first on Mobile World Live.
Login
