Samsung could start making its most important chip ever this month

Samsung will reportedly begin mass production of sixth-generation high-bandwidth memory (HBM4) chips this month. These advanced semiconductor memory chips are said to be used in Nvidia’s next-generation AI accelerator system, called Vera Rubin, which is expected to launch later this year.

A report from South Korea claims that Samsung could start shipping HBM4 chips to Nvidia as early as next week, coinciding with the Lunar New Year holiday. This lines up with an earlier report about the mass production schedule. Those chips will be used in Nvidia’s Rubin GPUs, which are designed for generative AI servers and hyperscalers. The company has recently invested a lot of resources to boost its HBM4 production capacity.

Samsung appears to be the first company to begin mass production of HBM4 chips. If this information is accurate, it may have surpassed SK Hynix, currently the world’s largest supplier of HBM chips. Samsung had been lagging behind SK Hynix in HBM3 and HBM3E chips, but it seems to have made significant progress with HBM4 and regained its competitive position.

Current-generation AI accelerators, such as the AMD MI350 and Nvidia B200, use fifth-generation high-bandwidth memory (HBM3E) chips. Micron and SK Hynix are the top two providers of HBM3E chips, while Samsung trails those firms. However, in the HBM4 segment, Samsung is expected to become the largest supplier.

Several reports indicate that Samsung’s HBM4 chips have passed Nvidia’s final verification testing, and Nvidia has already placed orders for them. Because Nvidia wanted early access to HBM4 chips, Samsung aligned its mass production schedule with Nvidia’s requirements.

Amazon and Google (Alphabet) could also use Samsung's HBM4 chips in their future AI accelerator systems.