Samsung to ship HBM4 chips to Nvidia in third week of February

Samsung Electronics is set to begin shipping the world's first mass-produced HBM4 chips to Nvidia in the third week of this month. The deliveries will follow the Lunar New Year holiday and support Nvidia's Vera Rubin AI accelerator platform.

Samsung Electronics plans to start shipping high-bandwidth memory 4 (HBM4) chips to Nvidia in the third week of February 2026, following the Lunar New Year holiday on February 17. This marks the world's first customer delivery of a mass-produced version, destined for Nvidia's next-generation Vera Rubin AI accelerator platform.

According to industry officials, Samsung has completed Nvidia's qualification process for HBM4 and aligned the delivery schedule with Nvidia's launch plans for Vera Rubin. Nvidia expects to demonstrate the platform at GTC 2026, from March 16 to 19. CEO Jensen Huang stated at CES 2026 last month that Vera Rubin chips are in full production, with availability anticipated in the second half of 2026.

Samsung is producing its HBM4 chips using the 1c process—its sixth-generation 10-nanometer-class DRAM technology—for the DRAM cell die, and a 4-nanometer foundry process for the base die. This enables data processing speeds of up to 11.7 gigabits per second (Gbps), surpassing the Joint Electron Device Engineering Council's (JEDEC) standard of 8 Gbps. Samsung declined to comment on customer matters but noted that its HBM4 was selected for Nvidia's platform after exceeding JEDEC performance standards.

Competition with rival SK hynix is set to intensify. SK hynix employs TSMC's 12-nanometer foundry process for the base die and its own fifth-generation 1b DRAM process, potentially supplying larger volumes due to more stable yields. Samsung aims for an early market foothold through performance advantages, though achieving stable yields with advanced processes will be crucial. An industry source praised Samsung, which holds the world's largest production capacity and broadest product lineup, for regaining technological edge by being the first to mass-produce the high-performing HBM4.

Relaterade artiklar

Nvidia CEO Jensen Huang at APEC Summit announcing 260,000 GPU deployment for South Korean AI factories, with partner executives and tech displays.
Bild genererad av AI

Nvidia plans 260,000 GPU deployment for South Korean AI factories

Rapporterad av AI Bild genererad av AI

Nvidia has announced plans to deploy up to 260,000 of its latest Blackwell GPUs in South Korea, partnering with the government and companies like Samsung Electronics, SK Group, Hyundai Motor Group, and Naver Cloud to build AI factories. The initiative will boost the country's AI computing capacity and offer significant opportunities for Samsung and SK hynix in supplying high-bandwidth memory chips. Nvidia CEO Jensen Huang praised South Korea's world-class memory technology during the APEC CEO Summit in Gyeongju.

Korean chipmakers Samsung Electronics and SK hynix are ramping up memory production to meet surging demand from artificial intelligence (AI) servers. Industry analysts said on Sunday that production capacity is increasingly seen as a key determinant of competitiveness in the global semiconductor industry.

Rapporterad av AI

Nvidia CEO Jensen Huang expressed confidence that South Korean companies will nimbly handle US tariff concerns during the Asia-Pacific Economic Cooperation (APEC) summit. He highlighted Korea's significant opportunities in physical AI and robotics, while praising partnerships with Samsung and SK hynix. The remarks came shortly after Seoul and Washington finalized a trade deal.

Ahead of CES 2026 in Las Vegas, major Korean tech firms including LG Electronics, Hyundai Motor Group, and Samsung Electronics unveiled AI-centric products and visions. They presented strategies like 'AI in Action' and 'Physical AI,' showcasing advances in robotics, laptops, memory, and more across daily life and industry. The events emphasized AI extending beyond screens into real-world applications.

Rapporterad av AI

Kina har godkänt import av Nvidias högpresterande AI-chips efter veckor av osäkerhet. Över 400 000 H200-chips är på väg till stora teknikföretag. Detta drag återspeglar Kinas strävan att möta teknologiska krav samtidigt som man eftersträvar självständighet.

President Donald Trump har meddelat en tull på 25% på vissa avancerade AI-chip från Nvidia och AMD, vilket tillåter export till Kina samtidigt som man kräver en andel av försäljningsintäkterna. Politiken vänder på ett tidigare exportförbud för Nvidias H200-chip men inför avgiften för att finansiera amerikanska intressen. Branschexperter ser det som ett sätt att skydda avtalet från rättsliga utmaningar.

Rapporterad av AI

Enligt Reuters har den kinesiska regeringen godkänt DeepSeeks köp av NVIDIAs avancerade H200 AI-chips. Detta godkännande kommer mitt i pågående amerikanska restriktioner och kinesiska företags ansträngningar att skaffa högpresterande hårdvara. ByteDance, Alibaba och Tencent har också fått tillstånd att köpa totalt 400 000 sådana GPU:er.

 

 

 

Denna webbplats använder cookies

Vi använder cookies för analys för att förbättra vår webbplats. Läs vår integritetspolicy för mer information.
Avböj