Copyright © BusinessAMBE 2023

Key takeaways

  • The Tesla AI5 chip will be produced by Samsung and TSMC in the US to meet Tesla’s growing demand for computing power in cars, robots, and data centers.
  • Tesla will continue to use Nvidia GPUs for model training alongside its own AI chips, in order to leverage the strengths of both technologies.
  • Tesla aims for higher performance per dollar with the AI5 by tailoring it specifically to the company’s needs, potentially achieving tenfold performance increases compared to other AI chips.

Elon Musk, CEO of Tesla, revealed details about the company’s next artificial intelligence chip, the AI5, during the Q3 earnings call on Wednesday. The AI5 will be produced by both Samsung in Texas and TSMC in Arizona, aiming for “overproduction” to satisfy Tesla’s needs in cars, robots, and data centers.

Tesla continues to use Nvidia GPUs

Despite developing its own chips, Tesla will continue to use Nvidia’s graphic processing units for model training due to their dominance in the AI market. Musk stated that Tesla wants to use both technologies together rather than fully replacing Nvidia.

Musk also pointed to Tesla’s substantial computing capacity, equivalent to 81,000 Nvidia H100 chips, highlighting the company’s commitment to enhancing its AI capabilities. The announcement follows the departure of former Apple engineer Peter Bannon, who previously led Tesla’s chip design and the development of the Dojo supercomputer designed to improve driverless technology.

AI5 chip

The AI5 chip, first announced in 2024, is the latest iteration of Tesla’s Autopilot hardware responsible for processing the signals necessary for autonomous driving. Earlier this year, Samsung closed a $16.5 billion chip deal with Tesla, confirming Musk’s statement that the AI5 would be produced by TSMC and its successor, the AI6, by Samsung.

Musk further clarified that both foundries would produce the AI5 at their US facilities, noting its smaller size compared to the “full reticle” designs of Nvidia and AMD.

Hyperscaler trend to develop custom AI chips

Tesla’s decision to design its own chips follows a trend among hyperscalers such as Google, Amazon, and Microsoft, who are developing custom AI chips for cost efficiency and performance optimization.

Musk emphasized the advantage for Tesla of being the sole customer for the AI5, allowing designers to tailor it specifically to the company’s needs and eliminate unnecessary components, potentially leading to superior performance per dollar. He predicted that the AI5 could be up to ten times more powerful than other artificial intelligence chips.

Musk acknowledged Nvidia’s success in meeting complex demands but expressed Tesla’s ambition for “radical simplicity” in chip design. Musk’s AI startup xAI, which collaborates with Tesla, has become a major Nvidia customer.

Follow Business AM on Google News as well

Want access to all articles? Take advantage of our summer promo and subscribe here!

© The Content Exchange, source News