OpenAI Models Advance Chip Design for Jalapeño and AMD
OpenAI is leveraging its frontier models to significantly improve chip design, enhancing its custom inference chip, Jalapeño, and collaborating with AMD for third-party hardware optimisation.

By Clara Tan, Artificial Intelligence
2 min read
OpenAI is pushing the boundaries of chip design by applying its advanced AI models to optimise hardware. The company announced on November 18, 2026, that its frontier models are being used to accelerate the development of its custom inference chip, Jalapeño, and to enhance third-party hardware in collaboration with AMD.
Early models, such as GPT-o3 mini used since early 2025, have already demonstrated significant improvements in Jalapeño's compute blocks. These improvements, measured in power, performance, and area (PPA), were observed in both the vector and matrix units. For instance, in the vector unit, the AI models identified opportunities to optimise floating-point adders, multipliers, and reduction logic by exploring alternative logic implementations and computational division across pipeline stages.
Similarly, in the matrix unit, the models contributed to improvements in floating-point adders, multipliers, and low-precision dot-product circuits. These optimizations highlight the AI's ability to find efficiencies that might be overlooked by traditional design methods.
OpenAI's collaboration with AMD extends these AI-driven design principles to external hardware, aiming to deliver substantial gains beyond their internal projects. Furthermore, OpenAI is partnering with Synopsys to integrate this frontier intelligence directly into electronic design automation (EDA) tools. This initiative seeks to empower chip engineers with AI capabilities within their existing workflows, streamlining the design and verification processes.
This development signifies a growing trend where AI is not just a software layer but a fundamental tool in the creation of the very hardware that powers it. The advancements promise more efficient and powerful chips, potentially accelerating the pace of innovation across the tech industry.
Sources
- 1OpenAI Blogopenai.com