Realtime AI News
OpenAI Unveils Jalapeño Custom Inference Chip Built with Broadcom, Moving Away from Nvidia
OpenAI has revealed plans for Jalapeño, its custom AI inference chip developed in partnership with Broadcom, joining a growing list of tech giants building their own silicon.
According to TechCrunch, OpenAI has shared plans for its custom AI inference chip, codenamed Jalapeño, built in collaboration with Broadcom. The move marks OpenAI's entry into the growing club of tech giants — including Google, Apple, and SpaceX — that are developing their own chips to reduce single-supplier dependence on Nvidia.
The Jalapeño chip is specifically designed for AI inference workloads rather than training, indicating OpenAI is building its own hardware foundation for large-scale inference deployments. This could significantly reduce its procurement dependency on Nvidia's GPUs, which have dominated the AI chip market for years.
The OpenAI-Broadcom partnership model resembles Google's collaboration with Broadcom on its TPU line — OpenAI defines the chip architecture requirements while Broadcom provides design services and IP integration, substantially shortening the development cycle.
This is widely seen as a landmark event for AI supply chain diversification. If Jalapeño meets its performance and cost targets, it could encourage more tech companies to follow suit, further shaking Nvidia's dominance in the AI inference chip market.
Why it matters
OpenAI's custom Jalapeño inference chip signals the accelerating trend of supply chain de-Nvidia-fication among AI giants, potentially reshaping the AI chip competitive landscape.