In the recent quarterly Tesla conference, CEO Elon Musk expressed his intent to enhance Tesla’s computing resources through both, NVIDIA accelerator purchases and by developing the company’s proprietary supercomputer, Dojo. Furthermore, he assured that Tesla is ready to buy AMD accelerators in addition to the NVIDIA equipment.
Prior to expressing interest in AMD accelerators, Musk hadannounced plans for Tesla to spend over $500 million on NVIDIA components this year. When asked on social media platform X, whether Tesla is open to also purchasing AMD components, Musk confirmed in the affirmative. It appears the interest is in the AMD Instinct MI300 accelerators currently released in the market.
The need for the development of Tesla’s proprietary AI system is increasing as the company recently began distributing a new version of the FSD v12 system among beta testing program participants. The system now relies on the operation of neural networks, not just for image recognition, but also for decision-making in vehicle control.
Notably, Musk disclosed that over $1 billion will be allocated towards the development of the Dojo supercomputer, underpinned by neural processors of Tesla’s own making, by the end of the present year. Musk stressed at the quarterly conference that Dojo is a long-term project with a high expected return on investment. His declaration considering the purchase of AMD chips prompted a temporary increase in the company’s stock price, which was later overtaken by a decrease. According to Musk, only around 10,000 NVIDIA H100 accelerators can be bought with $500 million and to remain competitive in the AI market, the company plans to spend several billion dollars annually on enhancing its computing power. $500 million will be spent solely on the development of the Dojo supercomputer in New York state this year.
This post was last modified on 01/27/2024