How the AI revolution is turning rivals into allies
Tesla CEO Elon Musk is doubling down on his company's use of Nvidia hardware to power its fleet of self-driving cars, with the company reportedly ordering 10,000 H100 chips for its AI cluster.
While Musk and Nvidia CEO Jensen Huang have not necessarily been friendly with each other in the past, recent developments show that their companies are doing business a lot more frequently.
Nvidia’s graphics processing units (GPUs) and other products and services have become very much in demand, especially with the boom of AI technologies. Its accelerator chips, in particular, are used to provide enough computing power for gen AI workloads, data centres, and robotics research.
READ MORE | How AI is redesigning talent development
Musk’s infatuation with Nvidia
Business is booming for Nvidia, so much so that the company recorded a US$22.1bn revenue in its latest quarterly report, which is equivalent to a 265% increase. In 2023, the global chip maker even surpassed its closest rival, Intel, in total sales.
Nvidia’s good fortune has caught the eye of Musk and Tesla, who are in the process of developing the company’s Dojo supercomputer. Dojo makes use of large amounts of video and data capture via Tesla vehicles to process and train AI models. Tesla is looking to use Nvidia GPUs to power more of its AI projects.
Musk is no stranger to developing artificial intelligence technologies, having been involved in the creation of OpenAI. The Tesla CEO was part of the AI research organisation before his public falling out with OpenAI CEO Sam Altman and other founders.
Now, Musk and his team are building their own powerful AI, a project that will require a significant amount of Nvidia hardware.
The close business relationship between Musk and Nvidia was visible during the chip maker’s recent GTC conference in California. The event had two sessions featuring representatives of xAI, the artificial intelligence start-up that Musk founded in 2023.
Igor Babuschkin, a co-founder of xAI and veteran of Open AI, gave a presentation of how they are leveraging the graphics processing units from Nvidia to develop Grok, which is xAI’s AI chatbot.
When Nvidia announced the launch of its Blackwell AI GPUs, Musk was quick to voice his support of the technology.
“There is currently nothing better than NVIDIA hardware for AI,” the Tesla CEO said.
READ MORE | How businesses can maximise AI for growth
Using Nvidia GPUs for Tesla AI cluster
Nvidia GPUs are an integral part of Tesla’s AI development. In fact, the company reportedly built its AI cluster using 10,000 H100 chips, according to former engineer Tim Zaman.
Musk appears to be very serious in his push for artificial intelligence, already spending $500 million to build the Dojo supercomputer. However, the billionaire entrepreneur made it clear that he is willing to shell out more money to develop Nvidia-powered AI.
“Tesla will spend more than that on Nvidia hardware this year,” Musk said.
“The table stakes for being competitive in AI are at least several billion dollars per year at this point.”
When asked whether Tesla was using Nvidia GPUs to train the Dojo supercomputer, Musk explained that they are “pursuing the dual path of Nvidia and Dojo”. This suggests that the company is developing Dojo without the use of Nvidia hardware. Instead, it is likely using the chip maker's technology for a different project/
“[Dojo] is a long shot worth taking because the payoff is potentially very high. But it’s not something that is a high probability,” Musk said.
During an earnings call, Oracle founder Larry Ellison revealed that the company helped xAI secure Nvidia GPUs for the first version of Grok. However, he said that Oracle was not able to meet the demands of the Tesla CEO.
“Boy, do they want a lot more GPUs than we gave them,” Ellison said. “We gave them quite a few, but they wanted more and we are in the process of getting them more.”