Serving technology enthusiasts for more than 25 years. TechSpot is a trusted source for tech advice and analysis.
In a nutshell Tesla is at the forefront of a heated battle between South Korean semiconductor giants Samsung & SK Hynix. The electric vehicle maker has reportedly contacted both companies to request samples of their upcoming HBM4 Memory Chips. Tesla
wants to integrate its Dojo system with the next-generation high-bandwidth memories. Dojo is a supercomputer that was built specifically to train “Full Self-Driving” the company’s neural networks. Insiders in the industry believe that Tesla could use the upgraded memory in Dojo as well as in its data centres and future self-driving cars.
The Dojo system currently uses older HBM2e chip to train the complex AI model that underpins Tesla’s Full-Self-Driving capabilities. As a TrendForce article citing Maeli Business Newspaper reveals, the company is looking to take advantage of the performance boosts that HBM4 promises.
High-bandwidth RAM, such as HBM4, is a specialized type designed to deliver high data throughput with greater energy efficiency. These qualities make it perfect for the processing power needed by cutting-edge AI workloads.
A third HBM4 innovation that is expected is an integrated logic chip acting as a memory stack controller. This change could unlock additional speed and power optimizations, ideal for AI data processors.
SK Hynix, Samsung and other major players are in a fierce competition to claim their share of the HBM market. This market is expected to grow to $33 billion in 2027. Both rivals are working hard to develop HBM4 prototypes for Tesla and other US tech giants such as Microsoft, Meta and Google.
SK is currently leading this race, supplying chips to Nvidia. It aims to start HBM4 production by late 2025. The company also took the lead by launching 321-layer NAND flash memory as the first. Samsung is determined to gain ground. It has partnered up with TSMC in order to produce key components using its advanced 4nm processing node.
