Nvidia has unveiled a new tech platform for self-driving cars as the world's leading chip-maker seeks more physical products to embed AI into.

Speaking at the annual CES technology conference in Las Vegas, boss Jensen Huan said the system - called Alpamayo - would bring "reasoning" to autonomous vehicles.

That would allow cars to "think through rare scenarios, drive safely in complex environments, and explain their driving decisions," Huang claimed.

He said Nvidia was working with Mercedes to produce a driverless car powered by the tech, which would be released in the US in the coming months before being rolled out in Europe and Asia.

Nvidia's chips have helped power the AI revolution, though so far attention has been chiefly focused on the software it powers, such as ChatGPT.

However, leading tech firms are now increasingly looking for hardware - meaning physical products such as cars - that AI could be used in.

Physical AI's 'ChatGPT moment'

Wearing his trademark black leather jacket, Huang told an audience of hundreds that the project has taught Nvidia "an enormous amount" about how to help partners build robotic systems.

"The ChatGPT moment for physical AI is almost here," Huang said.

"NVIDIA's pivot toward AI at scale and AI systems as differentiators will help keep it way ahead of rivals," said Paolo Pescatore, analyst at PP Foresight, from Las Vegas.

"Alpamayo represents a profound shift for NVIDIA, moving from being primarily a compute to a platform provider for physical AI ecosystems."

Shares of the AI chip designer rose slightly in after-hours trading following Huang's presentation.

It featured a video demonstration of the AI-powered Mercedes-Benz driving through San Francisco while a passenger, sat behind the steering wheel, kept their hands in their lap.

"It drives so naturally because it learned directly from human demonstrators," Huang said, "but in every single scenario... it tells you what it's going to do, and it reasons about what it's about to do."

Alpamayo is an open-source AI model, with the underlying code now available on machine learning platform Hugging Face, where autonomous vehicle researchers can access it for free and retrain the model, Huang said.

"Our vision is that someday, every single car, every single truck, will be autonomous," he told the audience.

The project could pose a threat to companies like Elon Musk's Tesla, which offers driver assistance software called Autopilot.

"Well that's just exactly what Tesla is doing," Musk posted on social media following the Alpamayo announcement. "What they will find is that it's easy to get to 99% and then super hard to solve the long tail of the distribution."

Like Tesla, Nvidiaalso has plans to launch a robotaxi service by next year in collaboration with a partner, but has declined to name the partner or say where it will be.

Nvidia is the world's most valuable publicly traded company, with a market cap of more than $4.5tn (£3.3tn).

It became the first company to reach $5tn in October, but has lost value over concerns about whether demand AI is overhyped.

The company also revealed that its Rubin AI chips are currently being manufactured and are due for release later this year.

The highly anticipated hardware can compute with less energy than Nvidia's current line of AI chips and could help lower the cost of developing the technology.