OpenAI and Broadcom launches Jalapeño, a custom chip built for AI inference

OpenAI launched its first custom-built Intelligence Processor called Jalapeño on June 24, in collaboration with Broadcom. The chip is primarily designed for Large Language Model (LLM) inference, which is the process of generating responses after a model has already been trained.

The processor is the first series of AI accelerators with a multi-generation hardware platform that could make AI systems faster, more reliable, and more accessible. 

Must read: Anthropic accuses Alibaba of largest-ever attempt to extract Claude AI capabilities

OpenAI’s custom AI inference chip

In a company blog post, OpenAI revealed that the chip was developed from the ground up, wherein the design was based on OpenAI's knowledge of LLMs, its future AI roadmap, software, and computing requirements of its products like ChatGPT. On the other hand, Broadcom and Celestica helped the company in chip engineering, implementation, and hardware infrastructure.

OpenAI highlighted that Jalapeño can run not only its own models but also on other LLMs. Early engineered versions of the chip are claimed to run AI workloads in testing labs, and its already said to have better performance-per-watt than current alternatives.

Hock Tan, President and CEO, Broadcom, said, “This is just the beginning of a multi-generation roadmap. By co-developing our industry-leading silicon directly with OpenAI, we are enabling the deployment of gigawatt-scale data centres with Microsoft and other partners beginning in 2026.”

The company wrote, “OpenAI is not only developing frontier models or building products on top of them; it is designing the infrastructure underneath them: chip architecture, kernels, memory systems, networking, scheduling, deployment systems, and product experience.”

Must read: Meta CPO reveals why CRED’s Kunal Shah was chosen to lead WhatsApp

The blog post further revealed how quickly OpenAI and Broadcom developed the Jalapeño chip. It highlighted that the chip went from the initial design stage to manufacturing in just nine months. The companies say that it may be the fastest development cycle ever for a high-performance custom semiconductor (ASIC).

OpenAI said that AI models, which are broadly used by consumers, are now helping engineers build the hardware needed for future AI systems. This has helped make chip development faster, cost-effective, and more efficient.

How can OpenAI Jalapeño be used?

Jalapeño chip is designed to run AI models after they have already been trained, making it a hardware platform to run consumer-centric AI applications. The chip is specifically optimised for AI inference, which lets AI models answer user prompts in ChatGPT, write code, or interact with an AI agent.

OpenAI believes that Jalapeño can help generate responses using less power, reduce hardware costs, and increase the number of users served per chip.