Strategic Memory Deal Boosts OpenAI’s AI Compute Roadmap
OpenAI announced that it has entered into letters of intent to secure high‑bandwidth memory (HBM) DRAM wafers from two of the world’s largest memory‑chip manufacturers, Samsung Electronics and SK Hynix. The agreements were finalized following a high‑level meeting in Seoul that included OpenAI chief executive Sam Altman, South Korea’s president Lee Jae‑myung, Samsung executive chairman Jay Y. Lee, and SK Hynix chairman Chey Tae‑won.
Under the terms of the deals, Samsung and SK Hynix will scale their manufacturing lines to produce up to 900,000 HBM DRAM chips per month for use in OpenAI’s Stargate AI infrastructure project and related data‑center deployments. SK Group noted that this output would more than double the current industry capacity for high‑bandwidth memory chips.
Stargate Project and Broader Compute Expansion
Stargate is a massive infrastructure initiative co‑led by OpenAI, Oracle, and SoftBank that aims to invest $500 billion to build data centers dedicated to AI development in the United States. The new memory supply agreements complement recent announcements that Nvidia will invest up to $100 billion in OpenAI, providing access to the ChatGPT maker to more than 10 gigawatts of compute capacity via Nvidia’s AI training systems.
OpenAI has also outlined plans to construct five data centers with SoftBank and Oracle as part of the Stargate effort, targeting a total compute capacity of 7 gigawatts. Earlier collaborations include Oracle’s commitment to sell $300 billion of compute capacity to OpenAI over five years, and a separate arrangement with SK Telecom to build an AI data center in South Korea.
Integration of OpenAI Services into Partner Operations
Beyond chip supply, Samsung and SK Hynix will embed OpenAI’s APIs and ChatGPT Enterprise into their internal workflows. This integration is expected to enhance the partners’ own AI capabilities while ensuring a close technological alignment with OpenAI’s platform.
OpenAI is also exploring additional opportunities with Samsung subsidiaries to expand AI data‑center construction outside Seoul, reflecting a broader strategy to diversify its geographic footprint for AI compute resources.
Implications for the AI Hardware Landscape
The agreements signal a concerted effort by OpenAI to secure critical hardware components that underpin large‑scale AI training models and applications. By locking in a substantial supply of HBM DRAM, OpenAI aims to mitigate potential bottlenecks in memory availability, a key constraint for high‑performance AI workloads.
For Samsung and SK Hynix, the deals represent a significant expansion of their high‑bandwidth memory production capabilities, positioning them as essential partners in the next wave of AI infrastructure development.
Looking Ahead
As OpenAI continues to scale its compute capacity through partnerships across the hardware and cloud ecosystems, the memory agreements with Samsung and SK Hynix will play a pivotal role in supporting the data‑intensive demands of its AI models. The collaboration underscores a growing trend of AI‑focused firms forging deep, multi‑year relationships with semiconductor manufacturers to secure the hardware foundations needed for future AI breakthroughs.
Este artigo foi escrito com a assistência de IA.
News Factory APP - notícias agênticas para impulsionar seu SEO e AEO.