Nvidia CEO Jensen Huang publicly mentioned foundry (contract chip manufacturing) cooperation with Samsung Electronics in his keynote address at Nvidia's annual event, "GTC 2026," drawing renewed attention to the partnership between Samsung Electronics' foundry division and Nvidia. After winning orders for Nvidia's Pascal graphics processing unit (GPU) in 2016, Samsung Electronics was not chosen by Nvidia for advanced process nodes thereafter.
However, with Huang officially saying that Samsung Electronics is producing the Groq 3 language processing unit (LPU) for artificial intelligence (AI) chips, the prevailing view is that he showed confidence in Samsung's foundry technology competitiveness. At this year's GTC, Huang met Han Jin-man, head of Samsung's foundry division, raising expectations that future order volumes could expand.
According to the industry on the 17th, Han Jin-man, president of Samsung Electronics, met CEO Huang at GTC 2026 and is said to have discussed future business cooperation. From Samsung Electronics' memory division, Hwang Sang-jun, head of memory development (executive vice president), joined the meeting. The fact that Hwang Sang-jun, a key executive in Samsung Electronics' high bandwidth memory (HBM) development, and Han Jin-man, who leads the foundry division, both attended GTC is seen as a move to emphasize that the company can provide not only memory semiconductors but also foundry services in an integrated offering as corporations.
A source familiar with Samsung said, "Among Nvidia's partners, the only supplier that does not rely on TSMC and can deliver HBM customized with advanced process foundry is currently Samsung Electronics," adding, "Given the characteristics of next-generation AI memory, which requires closer collaboration with foundries, Samsung's position in Nvidia's supply chain has become even more important."
In fact, CEO Jensen Huang highlighted Samsung Electronics as a core business partner in his keynote address that day. Huang said, "Samsung Electronics is producing the Groq 3 LPU for Nvidia. We are ramping up production as fast as possible, and we thank Samsung." He disclosed a specific timeline, saying the chip will begin shipping in the third quarter of this year. According to the industry, the Groq 3 LPU is being produced on Samsung Electronics' 4-nanometer process.
With Nvidia officially mentioning not only Samsung Electronics' HBM but also its foundry capabilities, many expect the scope of cooperation between the two companies to expand. In 2016, Samsung Electronics, leveraging the then state-of-the-art 14-nanometer FinFET process, jointly won orders for Nvidia's Pascal GPU along with TSMC. But results have been sparse in advanced process nodes since then. It won orders for the Orin automotive system-on-chip (SoC) and a chip for the Nintendo Switch 2, but both were 8-nanometer chips on an older process.
The industry interprets this as a sign that Samsung Electronics is regaining the trust of global heavyweights in advanced process nodes. In fact, Samsung Electronics' foundry division has been landing large-scale orders in succession from the industry's biggest customers, including Tesla. Tesla signed a $16.5 billion chip supply contract with Samsung Electronics in Jul. last year. Tesla CEO Elon Musk said Samsung's Taylor, Texas, plant will produce Tesla's next-generation AI6 chip. AMD and Qualcomm are also in talks with Samsung Electronics' foundry division.
Meanwhile, Samsung Electronics plans to sharply strengthen the profitability of its foundry division starting this year. It aims to rapidly expand its customer base in areas that require advanced process nodes, such as AI accelerators, data center semiconductors, and Autonomous Driving chips. In fact, Samsung Electronics is pursuing various collaborative projects, including test chips (MPW), with companies such as Broadcom, Intel, Cisco, and Infineon Technologies AG, as well as Tesla.