Ford Lead Says HPC, GPUs, AI Are Keys to Driving Progress

Like all auto manufacturers, Ford is pursuing what can only be described as a profound digital transformation.

At the SC24 conference, David Kepczynski, the carmaker’s global chief of digital transformation for product development, outlined how high-performance computing, GPU acceleration, and AI/ML are reshaping the company’s approach to everything from safety, performance, and connected vehicles.

It would be hard to find a better speaker for both exascale-class technologies and industrial uses of HPC. Before joining Ford, Kepczynski spent a decade at General Electric as CIO for Corporate Digital Technologies and Research, overseeing the development of large-scale computing systems for industries including healthcare, energy, and aerospace. And prior to that, his 20 years at General Motors included leadership roles in global digital engineering, product development, and manufacturing.

Kepczynski also co-chaired the U.S. Department of Energy’s Exascale Computing Project Industry and Agency Council, giving him a front-row seat to the evolution of HPC into what we see today—a mix of massive-scale GPU accelerated systems tuned for AI and/or traditional HPC as well as all the software footwork required at that scale.

While we were hoping for more in terms of architectural detail of Ford’s HPC environments, he did tell the SC24 crowd that their systems support 2,000 computer-aided engineers working on modeling and simulation (mostly CFD sims) daily.

Responding to a question from a Broadcom audience member about on-premises versus cloud HPC, Kepczynski explained that Ford relies primarily on on-prem systems for its modeling and simulation workloads. However, past experiments with hyperscale cloud providers such as AWS and Azure demonstrated the flexibility of hybrid approaches.

Kepczynski also explained that compute resources are evenly distributed among crashworthiness, aerodynamic performance, and connected vehicle analytics. However, he predicted a significant increase in AI/ML workloads in the coming years, particularly in connected vehicles, where real-time analytics are expected to dominate. “We anticipate that machine learning will account for up to 90% of compute cycles in some areas,” he said.

And while there was plenty of talk about how and where AI will go for Ford, he emphasized that good old fashioned HPC underpins every facet of vehicle development. In the realm of safety, for example, he says Ford has shifted from isolated crash simulations to a system-level approach, virtualizing occupants and crashes as interconnected components within a unified model. “We’re not just thinking about crashworthiness in isolation,” he said. “We’re leveraging HPC to integrate structural, mechanical, and software systems, delivering outcomes that enhance first-time quality and accelerate timelines.”

“This is about more than just transforming our vehicles. It’s about transforming how we think, design, and innovate as an industry.”

Performance optimization follows a similar systems-oriented philosophy. While aerodynamics remains a critical focus, Ford engineers now use HPC to integrate propulsion systems and NVH (noise, vibration, and harshness) characteristics into comprehensive vehicle simulations. This holistic approach, powered by advanced physics-based modeling, allows the company to fine-tune every aspect of vehicle performance. “We’ve moved beyond component-level analysis,” Kepczynski emphasized. “With HPC, we’re analyzing full systems, combining data, models, and compute power to achieve unprecedented fidelity.”

The blend of HPC, AI/ML, and GPU acceleration is coming into play right where we might most expect it: the connected vehicle. He describes how, by treating each car as an IoT platform, Ford integrates sensor data with real-time analytics to improve performance and reliability. These capabilities, powered, he says, predominantly by GPU-driven compute systems, enable advanced applications such as predictive maintenance and on-the-fly performance adjustments. “In this space, GPUs are essential,” Kepczynski noted. “They drive the large-scale data analytics that underpin our connected vehicle strategies.”

Kepczynski repeated the value of GPUs and work done in exascale computing. “Programmable GPUs changed everything,” he said. “Even with existing codebases, refactoring has unlocked extraordinary performance gains.”

As noted earlier, AI and machine learning are integral to Ford’s approach, with practical applications driving generative design, predictive analytics, and connected vehicle functionality.

Kepczynski dispelled any notions of hype, emphasizing that AI at Ford is focused on tangible outcomes. Surrogate models created through machine learning are integrated into design workflows, enabling faster iterations without compromising quality.

“We’re not just experimenting—we’re applying these technologies to deliver real results,” he said.