OpenAI Unveils Jalapeno Chip for Next-Generation AI Inference
OpenAI has unveiled Jalapeño, its first custom Intelligence Processor designed for large language model inference. Developed alongside Broadcom in a nine-month cycle, the chip promises superior performance per watt. It represents a core part of a multi-generation strategy to scale data centre capacity and make advanced AI more affordable for everyone
OpenAI has officially unveiled Jalapeño, its first custom-designed Intelligence Processor, marking a significant move toward building a full-stack infrastructure for artificial intelligence. Developed in close collaboration with Broadcom, the new accelerator is specifically engineered to handle the demands of current and future large language models (LLMs). According to a report by OpenAI, the chip has already demonstrated performance per watt figures that are substantially better than current state-of-the-art industry standards during early laboratory testing.
The development of Jalapeño was completed in a record-breaking nine-month cycle from initial design to production tape-out, a process significantly accelerated by utilising OpenAI’s own AI models. This chip is not a general-purpose processor but a dedicated inference engine designed from the ground up to reduce data movement and balance memory, networking, and compute resources. By controlling the hardware stack, OpenAI aims to ensure that its models, including iterations such as GPT-5.3-Codex-Spark, operate closer to their theoretical peak performance while maintaining higher efficiency for interactive AI services. OpenAI Introduces Lockdown Mode for ChatGPT Users.
OpenAI Optimising AI Infrastructure for Global Scale
The deployment of these accelerators is scheduled to begin in 2026, with plans to integrate the technology into gigawatt-scale data centres. This rollout is supported by a multi-generation partnership involving Broadcom, which provides essential silicon implementation and networking technology, and Celestica, which handles rack and system integration. By designing the infrastructure alongside the software kernels and serving patterns, the company intends to lower the cost of compute, ultimately making advanced AI tools more accessible and affordable for developers, enterprises, and individual users worldwide. ChatGPT Revamped Version: OpenAI Set to Transform Flagship AI Chatbot into 'Super App'.
The introduction of Jalapeño represents a strategic shift toward a full-stack approach where model capability and hardware efficiency reinforce one another. As compute efficiency improves through custom silicon, the company can facilitate more sophisticated model training and faster inference, which in turn leads to more capable products. This cyclical progress is designed to keep AI development sustainable, allowing for the reinvestment of resources into future generations of infrastructure that can solve increasingly complex problems across the global economy.
(The above story first appeared on LatestLY on Jun 24, 2026 08:07 PM IST. For more news and updates on politics, world, sports, entertainment and lifestyle, log on to our website latestly.com).
