Samsung And SK Hynix Are Vying For Tesla's Dojo Supercomputer Chip Orders

Tesla, a global leader in electric vehicles and artificial intelligence (AI), is making significant strides in the AI hardware space by pursuing next-generation High Bandwidth Memory (HBM4) chips from South Korean semiconductor giants Samsung and SK Hynix. This move underscores Tesla’s ambitions to enhance its custom-built Dojo supercomputer, a critical system designed to train neural networks for its Full Self-Driving (FSD) technology and other AI applications. The competition between Samsung and SK Hynix to secure Tesla’s business has intensified, with both companies vying to deliver cutting-edge HBM4 solutions.

Tesla’s Dojo Supercomputer and Its Role in AI Development

At the heart of Tesla’s AI initiatives lies the Dojo supercomputer, a proprietary system powered by Tesla’s D1 AI chip. Dojo is specifically designed to train neural networks that support Tesla’s FSD technology, which aims to enable fully autonomous driving. The supercomputer processes massive datasets to improve the accuracy and reliability of Tesla’s AI models. Currently, Dojo utilizes older HBM2e memory chips, which, while effective, are increasingly insufficient to meet the growing demands of Tesla’s AI workloads.