Sources say Samsung is speeding up the production of HBM memory chips for AI to compete with Hynix and TSMC

CTOnews.com, June 27 (Xinhua) in May this year, Kye Hyun Kyung, president of Samsung's device solutions department, admitted that Samsung's contract manufacturing technology lags behind TSMC. He also said that Samsung will overtake TSMC within five years. Now it is reported that Samsung is preparing to mass-produce high-bandwidth memory chips for AI applications this year.

According to foreign media KoreaTimes, Samsung plans to mass produce HBM chips for AI in the second half of 2023, but at present, their main goal is to catch up with SK Hynix, which has rapidly gained a leading position in the AI memory chip market.

CTOnews.com notes that SK Hynix has about 50 per cent of the HBM market in 2022, while Samsung has about 40 per cent. Meguiar accounts for the remaining 10%. However, the HBM market as a whole is small, accounting for only about 1 per cent of the total DRAM market.

Nevertheless, with the growth of the AI market, the demand for HBM solutions is expected to increase. Samsung now plans to catch up with SK Hynix and mass produce HBM3 chips to cope with market changes. AI memory chips are becoming more and more common, and high-bandwidth storage solutions are getting more and more attention.

Earlier, it was reported that Samsung has won AMD and Google as its customers, Samsung will make Google's Tensor 3 chip on its third-generation 4-nanometer process node, and the rumored Exynos 2400 SoC may also be made on 4-nanometer process.

▲ Google's Tensor 3 chip rendering, image source XDA