OpenAI Finalizes AI Chip Design — Production Set to Begin Soon

OpenAI is making significant progress in developing its own AI processor as part of its effort to decrease dependence on NVIDIA’s solutions. According to sources, the first prototype is nearing the final design stage and will be sent to TSMC for production in the coming months. The manufacturing process is expected to take approximately six months, with mass production potentially beginning in 2026. However, the success of the first prototype is not guaranteed—if issues arise, additional diagnostics and a repeat production cycle may be necessary.

The development is reportedly led by former Google engineer Richard Ho, with the team having doubled in size over the past few months to reach 40 members. Broadcom is also involved in the project, though the details of its role remain undisclosed. OpenAI’s first chip will be primarily designed for AI workloads, with the potential for expanded capabilities in future iterations. The processor will be manufactured using TSMC’s 3nm process and will feature a systolic array architecture along with HBM memory.

If testing proves successful, OpenAI could offer a viable alternative to NVIDIA’s solutions as early as this year. This would strengthen the company’s market position and enhance its leverage in negotiations with hardware suppliers