Could Tesla’s New Supercomputer Make it an Even Better AI Play?

Reports this week said that Tesla unveiled a new artificial intelligence computing cluster, and some say the move could also make the automaker a strong AI play. The new $300 million supercomputer is expected to use Nvidia hardware, despite also pitting Tesla against the tech company in the stock market.

Above: A Tesla Model 3 (Image: Casey Murphy / EVANNEX).

Oppenheimer analyst Rick Schafer said in a report on Tuesday that Tesla was debuting a $300 million AI computing cluster, as detailed by Barron’s. The supercomputer is expected to be used for Tesla’s Full Self-Driving (FSD) beta down the road, and Schafer described it as “more powerful than the world’s third highest-performing supercomputer.” Additionally, he says it will use 10,000 Nvidia H100 GPUs to power performance.

“Tesla is significantly enhancing its computing capabilities to train its full self-driving technology faster,” Schafer wrote in the report. “Tesla plans to spend over $2 billion on AI training in 2023 and another $2 billion in 2024.”

Tesla’s FSD beta and Autopilot systems use driver data to help train AI on driving. These systems have constantly been touted by CEO Elon Musk, who also expects self-driving vehicles to create “software-like margins” in the future. Since all Tesla vehicles are equipped with the ability to activate FSD, a future robotaxi could create ongoing software revenue with the flip of a switch, as many bullish analysts have echoed.

Currently, Tesla’s FSD is still in a beta mode and has a long way to go before becoming truly self-driving. In its current state, the FSD beta is at a Level 2 automation, requiring drivers to remain attentive and ready to regain control of the vehicle at any point. Level 3 automation would allow for some brief periods of inattention, while Level 4 and 5 represent completely self-driving automation for significant stretches of driving.

Musk live-streamed himself using v12 of the FSD beta on X over the weekend, though it’s still imperfect, as the drive showed. However, it’s pretty impressive what the system can do, and it makes it easy to imagine a future in which AI can safely pilot our cars for us. Musk has also noted in the past that training AI to drive cars is a lot harder than the now-common use of AI to develop a generative language model, such as ChatGPT.

At the time of writing during Tuesday trading hours, after news of the computing cluster broke, Tesla shares were valued at $253.58 (+$14.76), up 6.18 percent since the day’s market open.

===