The takeaway: While OpenAI and other AI startups continue to lose massive amounts of money on their "future-proof" services, Nvidia keeps selling the GPUs needed to power large-scale model training and inference workloads. Bottom line: the GeForce giant is well-positioned to ride the AI wave far longer than almost any other company in the stock market.
Nvidia and Samsung have announced a partnership to build a new "AI megafactory" featuring significant acceleration capabilities. The South Korean conglomerate plans to install a cluster of 50,000 GPUs designed by Nvidia, with the primary goal of enhancing chip manufacturing for mobile devices and robotics.
Samsung has yet to provide detailed information about the new facility, but the deal has been confirmed by multiple sources. According to Raymond Teh, Nvidia's senior vice president for the Asia-Pacific region, the company aims to support the South Korean government's ambitious plans for AI technology deployment.
Samsung's Jay Y Lee, Hyundai's Chung Euisun and Nvidia's Jensen Huang eat at a fried chicken restaurant in Seoul.
Prior to Teh's confirmation, Nvidia CEO Jensen Huang was reportedly seen meeting with Samsung Chairman Lee Jae-yong and Hyundai Chairman Chung Eui-sun over beer. Nvidia is also collaborating with other major South Korean corporations to supply a comparable number of GPUs for accelerating AI workloads.
The collaboration between Nvidia and Samsung is expected to have a lasting impact on both companies, and potentially the chip industry as a whole. Samsung is one of the few companies worldwide with its own chipmaking facilities, while Nvidia continues to rely on third-party foundries to manufacture its complex, proprietary GPU designs for the market.
Nvidia will also adapt its designs to Samsung's lithography-based manufacturing platform for its new accelerators. According to Nvidia representatives, this closer technological cooperation could provide the Korean foundry with a 20 percent performance uplift. Nvidia's Omniverse platform, which integrates GPU technologies into existing software tools, is expected to play a role in the partnership as well.
Through this collaboration, Samsung strengthens its position as a key partner in Nvidia's high-performing chip ecosystem, while Nvidia secures its business prospects for years to come. The company recently became the first $5 trillion public company in Wall Street history, and Huang confirmed that partners have already booked $500 billion worth of GPU orders for both the Blackwell and upcoming Rubin GPU architectures.
Image credit: APEC 2025 Korea & Yonhap News
Nvidia and Samsung team up to build an AI megafactory with 50,000 GPUs

