Samsung to increase its HBM production capacity drastically amid rising AI demand

Around 2022, GenAI chatbots, such as ChatGPT, started hitting the market, and since then, the demand for AI has been skyrocketing. These tools run on servers equipped with AI accelerators, which use a large amount of high-bandwidth memory (HBM). As such, the demand for this type of memory also started rising sharply, increasing its prices, and bringing memory manufacturers a lot of money. 

Back then, Samsung used to make HBM chips, but it couldn’t sell many, as major brands consuming it, such as Nvidia, hadn’t approved these chips due to their lower performance than other HBM chips in terms of market. As Samsung was missing the opportunity to earn a lot of money, it made some major changes to the memory division and improved its HBM chips drastically. 

With that, Samsung’s HBM3E and HBM4 chips are in immense demand now. In fact, the latter has been found to be the best in the market by Broadcom and Nvidia. Well, there’s so much demand for the brand’s HBM chips that it isn’t able to meet the demand, which means that the company is losing the opportunity to earn even more money. To that extent, Samsung seems to be expanding the HBM manufacturing capacity.

According to a report from ET News, Samsung plans to expand its HBM production capacity by 50% by the end of 2026. Currently, the company makes around 170,000 wafers of HBM chips every month, which is similar to SK Hynix’s current HBM production. By the end of the next year, the tech giant is expected to make 250,000 wafers per month. The increased capacity is said to be focused mainly on HBM4.

The publication says that to increase the HBM production, Samsung will be investing money to expand the Pyeongtaek 4 (P4) facility.