OpenAI ropes in Samsung, SK Hynix to source memory chips for Stargate

OpenAI has partnered with Samsung Electronics and SK Hynix to significantly boost its AI compute capacity for the Stargate project, securing agreements for the production of up to 900,000 high-bandwidth memory (HBM) DRAM chips monthly, which more than doubles current industry capacity. This collaboration, part of OpenAI's $500 billion Stargate initiative with Oracle and SoftBank, also includes building AI data centers in South Korea and integrating OpenAI's APIs into Samsung and SK Group operations, underscoring the immense capital deployment and robust demand for specialized semiconductor manufacturing and data center infrastructure driving advanced AI development.

Analysis

OpenAI is executing an aggressive, large-scale strategy to secure critical components for its AI compute infrastructure, as evidenced by its new agreements with Samsung Electronics and SK Hynix. The deal secures the production of up to 900,000 high-bandwidth memory (HBM) DRAM chips per month, a figure projected to more than double the current global industry capacity for this essential AI hardware. This move directly addresses a key bottleneck in the AI supply chain and is a core component of the $500 billion Stargate infrastructure project, a joint initiative with Oracle and SoftBank. The frenetic pace of investment is further highlighted by Nvidia's recent commitment of up to $100 billion for access to over 10 gigawatts of compute capacity and a separate $300 billion, five-year compute sale agreement with Oracle. The strategy also includes significant geographic diversification, with plans to build multiple AI data centers in South Korea in partnership with SK Telecom and Samsung subsidiaries. As part of the arrangement, Samsung and SK Group will also become enterprise customers by integrating OpenAI's APIs, creating a powerful ecosystem of supply and demand among the key players.

Key Decisions for Investors

• The unprecedented capital commitments from OpenAI, Oracle (ORCL), and Nvidia (NVDA) confirm a durable, high-growth demand cycle for the entire AI hardware supply chain, supporting a bullish outlook on key component and infrastructure providers.
• Investors should recognize high-bandwidth memory (HBM) as a critical, capacity-constrained component, positioning HBM manufacturers like Samsung and SK Hynix as prime beneficiaries of the current AI infrastructure super-cycle.
• The expansion into South Korea for data center construction with partners like SK Telecom (SKM) indicates that geographically diversified infrastructure and telecommunication partners are becoming integral to the AI value chain.