OpenAI to Outsource Next-Gen AI Processor Production to Samsung Foundry, Alongside TSMC

OpenAI Expands Chip Manufacturing Partnerships

OpenAI is extending its collaboration with Samsung to include the manufacturing of its next-generation AI processors. Harrison Kim, General Manager of OpenAI Korea, confirmed that Samsung Foundry will produce at least some of these chips. This move signifies a broader relationship with Samsung, moving beyond memory supply and enterprise software.

Dual-Sourcing Strategy for AI Accelerators

OpenAI currently relies on TSMC for the production of its first inference AI accelerator, Jalapeño, which was co-designed with Broadcom. The decision to also utilize Samsung Foundry for future chips indicates a strategy to double-source its AI accelerators. This approach suggests that OpenAI anticipates massive volume requirements for its proprietary AI hardware.

Focus on Next-Generation Chips

While OpenAI's first-generation Jalapeño chip is already in mass production at TSMC, the partnership with Samsung is specifically for 'next-generation chips.' This implies that Samsung Foundry will likely be involved in the production of the successor to Jalapeño, which is currently under development and approaching tapeout. The exact nature of Samsung's involvement, whether for a new chip or a jointly developed one, was not specified.

Implications for AI Hardware Production

This development highlights OpenAI's commitment to developing its own AI ASIC program and securing diverse manufacturing capabilities. The need for multiple foundry partners underscores the increasing demand for specialized AI hardware and the scale at which AI companies are operating.