Elon Musk might aspire to build a supercomputer that could rival Nvidia, but his companies are getting cozier with the chipmaker, with Tesla and XAI on an Nvidia hardware-buying spree.
Nvidia CEO Jensen Huang offered a deftly diplomatic yet firm response when asked during an all-hands meeting if the company would follow Apple and Disney in pausing ads on Twitter over concerns about a recent surge in antisemitic and other hateful content on the platform. Huang made it clear that Nvidia hadn't advertised on Twitter in a long while and had no plans to start. But he pointedly declined to comment on other companies' decisions.
Conspicuously absent from Huang's remarks: any mention of Nvidia's deepening ties to Musk's business empire.
Amid surging demand for NVIDIA's graphics processing units (GPUs) and related hardware and services, the company's market capitalization has surpassed the $2 trillion mark. NVIDIA's chips power generative AI, robotics, scientific research, and data center projects.
NVIDIA's revenue jumped 265% to $22.1 billion in a recent quarter, outpacing industry giant Intel. Meanwhile, Elon Musk's ventures have been heavily acquiring NVIDIA hardware to bolster their ambitious AI development plans.
The close ties between NVIDIA and Musk's companies were on display at NVIDIA's annual developer conference in San Jose, California, this week. The event drew about 160,000 attendees, including celebrities and industry leaders. Executives from xAI, Musk's AI startup founded in July 2023, participated in two sessions.
xAI co-founder and research scientist Christian Szegedy, formerly of Google, joined NVIDIA data scientist Bojan Tunguz in a talk about using NVIDIA GPUs to accelerate the training and inference of xAI's conversational AI, Grok. Igor Babuschkin, xAI's other co-founder and a research engineer with a background at OpenAI and Google, shared how the startup uses NVIDIA technology to speed up Grok's training and enhance its reasoning capabilities.
During an event introducing NVIDIA's next-generation AI chip, Blackwell, Musk heaped praise on NVIDIA's technology. "There's really no AI hardware that compares to NVIDIA in the market today," he said.
Musk, who was briefly a founder of OpenAI, left the organization due to disagreements with other founders and started xAI to pursue a different approach to building AI models and software.
Meanwhile, Musk's electric car company, Tesla, has been aggressively developing AI software for its vehicles' autonomous driving capabilities. Tesla is also working on the Optimus humanoid robot, designed to perform complex tasks.
Although Tesla generates the majority of its revenue from car sales, Musk often urges shareholders to think of the company differently. "Most people think of Tesla as an auto company, but really it's an AI and robotics company," he wrote in a post on the X platform in January.
Tesla unveiled plans at its AI Day event in August 2021 to build the "Dojo" supercomputer, which will leverage the massive amounts of video and data collected by Tesla's fleet of cars to train AI models.
NVIDIA hardware plays a central role in Tesla's AI efforts. Tim Zaman, a former Tesla AI engineer, revealed on the X platform in August that a cluster of 10,000 NVIDIA H100 chips had been installed and was about to be brought online.
Despite the $500 million cost of building the Dojo supercomputer, Musk said in a January post that "Tesla's spending on NVIDIA hardware this year will be an order of magnitude more." He added that "to be competitive in AI, you really need to spend several billion dollars per year."
During Tesla's fourth-quarter earnings call in January, an analyst asked about the latest on Dojo and whether Tesla had "enough NVIDIA GPUs to train the system." Musk responded that Tesla was pursuing both the NVIDIA and Dojo paths, suggesting Tesla won't be fully reliant on NVIDIA technology for Dojo. He called Dojo a "long shot," but added that "the potential payoffs are so large that it's worth trying."
Musk's close friend and former Tesla board member, X investor, and Oracle founder Larry Ellison said during a company earnings call in December that when xAI was developing its first version of Grok, Musk needed more NVIDIA GPUs than Oracle could supply. "Boy, they wanted a lot more GPUs than we had access to. We provided a bunch, but they wanted even more," Ellison said at the time. "We're scrambling to make sure that we satisfy their demand for GPUs."
Despite Musk's long-standing connection to NVIDIA CEO Jensen Huang and the closer-than-ever partnership between their companies, the relationship hasn't always been smooth sailing.
"When Elon saw the machine, his response was, 'I want one.' He's the one who introduced me to OpenAI. I delivered the world's first AI supercomputer to OpenAI that day."
Nvidia spokesmen declined to comment, while Tesla and xAI did not respond to requests for comment.
