Samsung Electronics Co. and SK Hynix Inc. are said to each be developing a sixth-generation high-bandwidth memory (HBM4) chip prototype for Tesla Inc., which has joined its US Big Tech peers in a race to develop its own artificial intelligence chips, according to semiconductor industry sources on Tuesday.
A slew of next-generation HBM chip orders received by the South Korean memory chip cross-town rivals suggest the AI-driven HBM boom will continue through next year.
Industry sources said that the US EV giant has asked the Korean chip duo to supply HBM4 chips for general use, and it is expected to choose one of the two companies as its HBM4 supplier after testing their samples.
The Korean chipmakers have been developing customized HBM4 chips for US Big Tech companies such as Google LLC, Meta Platforms Inc. and Microsoft Corp., seeking to lower their reliance on Nvidia Corp.'s AI chips.
Joining the Big tech companies, Tesla is expected to use the next-generation HBM chip to enhance its AI chip capability.
Tesla operates Dojo, its custom-built supercomputer designed to train its “Full Self-Driving” neural networks. This is also expected to be the cornerstone of Tesla’s AI ambitions beyond self-driving.
HBM chips are one of the key parts in running the supercomputer to train AI models with massive datasets, and Tesla is expected to use the sixth-generation HBM chip in Dojo, also powered by its own AI chip D1.
The HBM4 chip could also be used in Tesla’s AI data centers under development and its self-driving cars, which are currently fitted with HBM2E chips for pilot programs.
In addition to Tesla, Samsung Electronics and SK Hynix are also developing customized HBM4 chips for large American tech companies such as Google, Meta, and Microsoft, which have been seeking to reduce their reliance on Nvidia AI chips.
The production of Samsung Electronics HBM4 chips is expected to begin in the fourth quarter of this year, and the test products are expected to be released around early next year. After verifying the product's operation, design and process improvements are also expected. The goal is to complete the preparation work for mass production of the 12 layer HBM4 before June next year. Samsung plans to use 1c DRAM as the HBM4 core chip and mass produce logic chips through foundry 4nm process. Therefore, Samsung Electronics must improve the mass production yield of 1c DRAM to meet the internal production schedule of HBM4.
SK Hynix plans to mass produce 12 layer HBM4 in the second half of next year, with a core chip using 1b DRAM, and logic chips produced using TSMC 5nm and 12nm processes. Recently, there have been reports that SK Group Chairman Choi Tae won revealed that Nvidia CEO has requested SK Hynix to advance the supply of HBM4 chips by six months. The CEO of SK Hynix stated that it is possible to achieve this. It is currently unclear whether this will change SK Hynix's previously announced production schedule.
|