Jensen Huang, the CEO of Nvidia, expressed his gratitude to Samsung during the keynote speech of the annual developers’ conference ‘GTC 2026’.
On March 16th, local time, Jensen Huang delivered a keynote speech at the SAP Center in San Jose, California, where he introduced a dedicated inference chip. He stated, “Samsung is manufacturing the ‘Groq3’ language processing unit (LPU) chip for us,” and added, “We are increasing production as quickly as possible. Thank you very much to Samsung.”
Jensen Huang further explained that the chip would be integrated into Nvidia’s next-generation AI chip ‘Vera Rubin’ system, noting, “The shipment is expected to start in the second half of this year, probably around the third quarter.”
The Groq3 LPU is a chip designed to enhance inference performance and efficiency by sharing roles with Nvidia’s ‘Rubin’ graphics processing unit (GPU). It was confirmed through Jensen Huang’s remarks that it is being produced by Samsung Electronics’ semiconductor foundry division.
Samsung Electronics also highlighted its collaboration with Nvidia in the memory sector by publicly unveiling the physical chip of the next-generation HBM, ‘HBM4E,’ and the ‘core die’ wafer for stacking at the GTC event exhibition hall.
The HBM4E, aiming for sample shipment in the second half of this year, is expected to support a data transfer rate of 16Gbps per pin and a bandwidth of 4.0TB/s.
[Photo Source: AP=Yonhap News]
■ How to Provide Information
▷ Phone: 02-781-1234, 4444
▷ Email: kbs1234@kbs.co.kr
▷ KakaoTalk: Search ‘KBS제보’, add the channel
▷ Subscribe to KBS News on YouTube and Daum too!