Tesla has announced an ambitious plan to significantly expand its capabilities in artificial intelligence (AI), with the installation of a cluster of 50,000 GPUs and increased investment in its customized supercomputer Dojo. This strategy aims to enhance the development of full self-driving (FSD) and other AI initiatives of the company.
Expansion of Technological Infrastructure
The announcement comes at a time when Tesla’s stock has dropped more than eight percent in pre-market trading, after reporting second-quarter revenue and profits below expectations, a 45 percent decrease in earnings, and a delay in the launch event of its ‘robotaxi’.
Despite a reduction in capital expenditure during the second quarter, Tesla’s CFO, Vaibhav Taneja, indicated that the company expects capital expenditures to exceed 9 billion euros. “The increase in our spending to bring online a cluster of 50,000 GPUs will greatly enhance our capabilities to scale full self-driving and other AI initiatives,” Taneja stated.
Development at the Gigafactory Texas
Tesla CEO, Elon Musk, mentioned that the company has been working tirelessly to complete the southern extension of the Gigafactory in Texas, where the 50,000 H100 GPUs will be housed. “That southern extension will house the 50,000 H100s, and we are starting to move the H100 server racks there,” Musk explained.
In response to questions about internal Nvidia emails revealing plans to send thousands of GPUs reserved for Tesla to its private X and xAI companies, Musk clarified that “Tesla’s data centers were full” and that it was in Tesla’s best interest to use the GPUs where they could be operational immediately.
Focus on the Dojo Supercomputer
During the earnings call, Musk emphasized the importance of Dojo, the supercomputer dedicated to FSD with custom D1 chips and unique system configurations. “The demand for Nvidia hardware is so high that it is often difficult to get the GPUs,” Musk commented, noting that this forces Tesla to step up efforts on Dojo to ensure the necessary training capacity.
“We are going to double down on Dojo and see a path to be competitive with Nvidia. We have no other choice due to the high demand for Nvidia and the elevated GPU prices,” Musk added.
Integration of AI Technology
On the X social network, Musk also mentioned that Tesla uses the HW4 AI computer in the training cycle with Nvidia GPUs, at a ratio of 1:2, and announced a name change from Hardware 4 (HW4) to Artificial Intelligence 4 (AI4). In addition, Musk indicated that “Dojo 1 will have approximately the equivalent of 8,000 H100s online by the end of the year.”
Lastly, Musk launched a poll on X asking users if Tesla should invest 4.5 billion euros in its xAI AI startup, highlighting that board approval and shareholder votes would be needed to proceed.
With these developments, Tesla reaffirms its commitment to expanding its technological capabilities, positioning itself at the forefront of the artificial intelligence and autonomous driving revolution.