A great acceleration is coming to the auto industry thanks to artificial intelligence, and one of the giants of the carmaking world is using the technology to reinvent its product pipeline.
In recent years, General Motors has dabbled with AI and virtual simulation tools to evolve the way it designs, engineers, tests, and manufactures its vehicles. Now the company has fully embraced these tools and woven them through the entire product development process. As a result, GM’s use of AI tools, generative design technology, and advanced simulations are cutting the typical car development timeline in half.
“Today, the average vehicle program probably across the industry, at least in the United States and Europe, is on the order of four to five years,” says GM’s chief product officer, Sterling Anderson. “Our target is two.”
Anderson left Aurora, the autonomous trucking company he cofounded, to join GM in 2025. Since then, he’s pushed the legacy automaker to take comprehensive advantage of AI, machine learning, modern computation, and generative optimization, unleashing these tools on every part of the car development process. From the earliest stages of concept design all the way through the tooling for manufacturing, GM now has a unified digital model that brings once-disparate disciplines onto the same virtual page.
Anderson calls it “a new operating system for product development.” It could also be a major strategic advantage for the company, which had $185 billion in revenue in 2025, a slight dip from 2024 but part of an upward trend in the past five years.
Central to this approach is a shift from sequential development to concurrent workflows. Rather than an engineering team waiting for a design team to iron out a car’s body contours before figuring out its structure, the design-in-progress can be used to build out the rough shape of the structure even as the overall form of the car evolves. Tweaks to one area—be it the aerodynamics, the crash safety systems, or the size of the chassis—can be shared immediately across the different disciplines developing the car and integrated into their own side of the project.
The change will have implications for design, engineering, validation and testing, and manufacturing. Virtual wind tunnel tests can happen simultaneously with car exterior designs. Co-simulations of different systems can manage energy use while optimizing interior cabin cooling. Virtual crash tests can inform vehicle chassis design even before a prototype is built.
“By compiling all of this into a single model, into a single optimization, we’re able to get a much more globally optimal answer in much less time,” says Anderson. “As a designer, when I can just tweak a surface and see immediately what the impact that just had on my coefficient of drag, I become a better designer. I certainly can iterate a lot faster than I used to be able to.”
GM’s deep pool of test data is central to this new approach, according to Jason Fischer, GM’s executive director of virtual integration engineering. During a recent call, he demonstrated a virtual model of a common vehicle test known as an avoidance maneuver. On his screen, a digital version of an SUV drives down a road and then abruptly swerves, as if to avoid a sprinting squirrel or another car veering across the lane divider. These tests are used to make sure a vehicle’s suspension can handle such a sharp but common maneuver, and the data collected in the tests are used to inform material choices, the design of parts, and the systems engineering that helps the various parts of a car talk to each other.
Within this virtual simulation and its large foundation of data pulled from real and virtual tests over the years, GM’s engineers can use AI to infer other avoidance maneuvers that could put strain on a part or an entire design.
“We can start expanding the number of scenarios that we do this in. We can change road conditions, load conditions, and dynamic events, and what you get out of that are results that don’t just perform well in ideal conditions,” Fischer says.
The simulation can then inform the rest of the development process, making the overall product better prepared for a wide range of on-road conditions. “It’s actually hardened against all the real-world physical potential situations that you as a customer would see,” he says.
This extends into crash testing, with AI-informed virtual crash data helping designers and engineers figure out what is unsafe long before turning a design into an expensive prototype. “Engineers find weak spots, they fix them faster, and they arrive at the physical testing in a stronger, more refined vehicle,” says Fischer. “We can make decisions in real time with artificial intelligence because the data is there.”
These approaches are already changing the way the company develops new products. GM’s approach using AI tools and virtual design is changing cars that are now coming to market. One example is the new Chevrolet Silverado launching later this year, which has a V8 engine that was designed using thousands of combustion chamber simulations. These computational designs led to GM’s engineers hitting performance and emissions goals in a third of the time they’d typically spend. It also helped cut the costs of producing physical prototype engines by 20%.
The average car buyer may not notice much change at first. But because GM’s AI tools are drastically increasing the speed at which the company can develop products, consumers will be seeing bigger changes from model year to model year, and likely new concepts turning into production vehicles much faster than before.
“When you can do this, you can build a broader portfolio that suits a larger range of needs with much higher refresh cycles, which is generally a good thing,” Anderson says. “The next generation of any vehicle is typically better than the last. And when you can cut that period in between more than in half, you’re getting just better vehicles evolutionarily.”