[Edaily Reporter Song Jae-min] Samsung Electronics is working to revitalize its foundry business by expanding its collaboration with NVIDIA on next-generation artificial intelligence (AI) semiconductors. Industry attention is focused on the matter after Jeon Young-hyun, Vice Chairman and Head of Samsung Electronics’ Device Solutions (DS) Division, directly mentioned cooperation on the next-generation version of the language processing unit (LPU) “Groq” immediately following a meeting with Jensen Huang, CEO of NVIDIA.

According to industry sources on the 9th, Vice Chairman Jeon met with reporters following an NVIDIA event held at the Shilla Hotel’s Yeongbin Hall in Seoul the previous day and stated, “We are currently collaborating on autonomous driving chips and Groq chips using 4nm and 8nm processes, and we are also discussing cooperation for the next generation.”

The industry views these remarks as significant, as this is effectively the first time Samsung Electronics has officially mentioned collaboration with NVIDIA on follow-up products to Grok. Samsung Foundry is currently known to be involved in the production of NVIDIA’s autonomous driving platform, “Drive AGX Thor,” and the Grok LP30.

In particular, Grok represents a key AI semiconductor order secured by Samsung Electronics as a core client for its Taylor, Texas, plant. However, since NVIDIA acquired Grok late last year, there have been persistent speculations that production of next-generation products could shift to TSMC’s 2-nanometer process. In fact, TSMC drew industry attention during its first-quarter earnings call this year by revealing that it is collaborating with a specific client on the development of next-generation LPUs.

Given this situation, the fact that the former vice chairman personally mentioned “next-generation cooperation” is interpreted as suggesting that Samsung Electronics still maintains a strong position in the competition to secure orders for the next generation of Grok.

Vice Chairman Jeon also mentioned the possibility of joint development beyond HBM collaboration. He stated, “In the short term, we discussed how to proceed with HBM4 and foundry collaboration, and in the medium to long term, we also discussed working together on joint development.” This suggests the potential to expand cooperation beyond a simple supply relationship to include collaboration starting from the development stage of next-generation AI semiconductors.

These remarks are also significant in light of the ongoing losses in Samsung’s foundry business. Samsung Electronics’ foundry division is reported to have recorded losses for five consecutive quarters through the first quarter of this year. However, the mood has been shifting since the second half of last year.

Following the signing of an approximately 23 trillion won AI chip supply contract with Tesla last year, Samsung Electronics has also secured production volume for Apple’s image sensors (CIS). Production of Grok chips for NVIDIA is also contributing to the recovery of utilization rates. CEO Jensen Huang also mentioned directly at GTC 2026 last March that “the Grok 3 LPU is being manufactured using Samsung Electronics’ 4-nanometer process.”

HBM4 is also cited as a new growth driver for the foundry business. This is because the application of Samsung Electronics’ 4-nanometer process to HBM4 base dies has created a structure where expanded HBM production directly leads to higher foundry utilization rates. Samsung Electronics is also accelerating efforts to secure customers for its 2-nanometer process.

The securities industry is also paying close attention to the potential for improved earnings. Kim Rok-ho, an analyst at Hana Securities, stated, “With yield rates stabilizing on the 4-nanometer process and shipments of Grok and HBM base dies for NVIDIA increasing, we expect earnings to normalize as utilization rates recover,” adding, “The foundry division is likely to return to profitability in the second half of this year.”