Ford: GPU compute in the cloud has been crazy expensive and also just difficult to get

"Bryan Goodman, Ford Motor's director of artificial intelligence, said he is pushing forward a number of use cases for artificial intelligence at the company, as part of an effort to design and engineer new cars as quickly as its competitors.

"I personally think that Ford engineers as good of products as anyone in the world in automotive. But I think we have to speed up to be competitive. I look at the Chinese [original equipment manufacturers] and their engineering cycle times are quite fast," he said.

Ford in recent months has grappled with timid demand for electric vehicles, rising competition and, more recently, new U.S. tariffs. Last month, the company posted higher revenue for the fourth quarter, but the outlook for 2025 was a big disappointment, causing shares to fall.

Goodman sat down Tuesday with The Wall Street Journal at Nvidia GTC in San Jose, Calif., an event he has attended for the past 10 years, to discuss his vision for implementing AI, his concerns about coming types of GPUs and more.

The interview has been edited and condensed for length and clarity.

WSJ: What role do you see AI agents playing in the vehicle design process?

GOODMAN: The designers sculpt 3D vehicles in this red clay, but it's very time consuming. They like to sketch, but that's two dimensional. By using AI, we can turn [those sketches] into three-dimensional models and renderings and connect that with engineering.

And we've trained AI systems to predict stresses and lots of different physics-based tests that we do in designing and engineering and testing a vehicle, like computational fluid dynamics, wind tunnel drag. One run would take 15 hours. But we've trained an AI model to predict what that would do and it runs in 10 seconds.

WSJ: What AI models is this using?

GOODMAN: We use OpenAI models, Google models, Anthropic models and open source models. We've used Meta's and some of the Chinese models and Mistral.

WSJ: You've used DeepSeek?

GOODMAN: We have, yes. We have found it's very good. But I don't see it as being any better than the latest Anthropic and [Google] Gemini and OpenAI models. No better or worse. The really nice thing, though, is that it is open source. And because they shared everything with the world, I think we're going to see a lot built on top of it.

WSJ: Where is it all running?

GOODMAN: We run most of that on our own GPUs. We have Nvidia GPUs. We have a few thousand. Five, six years ago, we moved everything to the cloud, except high-performance compute.

GPU compute in the cloud has been crazy expensive and also just difficult to get. I don't blame them for charging that much. Sometimes it just wasn't available.

WSJ: Nvidia's Blackwell chips were a main focus of last year's GTC event, but were then delayed due to manufacturing challenges. Do you have your hands on them yet?

GOODMAN: We do not have Blackwell yet. Nvidia officially says they are shipping, but we don't have any yet.

WSJ: Nvidia CEO Jensen Huang mentioned during his keynote that Blackwell is making the previous generation's Hopper chips essentially useless. What do you make of that?

GOODMAN: I don't quite see it as they're useless. He was, of course, trying to say that his new product is so much better than his old product, that he's making his old product obsolete. But I suspect we'll get a lot of use out of our Hopper GPUs the next few years. And we're working on Blackwell.” [1]

1.  Ford's Innovation Path: AI Agents, Nvidia GPUs. Bousquette, Isabelle.  Wall Street Journal, Eastern edition; New York, N.Y.. 21 Mar 2025: B4.