OpenAI and Broadcom Launch Jalapeño, a Revolutionary AI Chip

OpenAI and Broadcom Launch Jalapeño, a Revolutionary AI Chip

OpenAI unveiled its very first in-house AI chip, named Jalapeño, on June 24. Produced by the American semiconductor giant Broadcom, this chip is primarily intended for inference tasks and is expected to begin deployment by the end of 2026 for rapid operational use.

In the sweltering late June heat, OpenAI introduced "Jalapeño," its first custom chip dedicated to AI. Manufactured by Broadcom, this processor is specifically designed to meet OpenAI's needs and will primarily be used for inference tasks.

OpenAI Spices Up Its Strategy

Designed from the ground up by OpenAI engineers, with support from those at Broadcom and Celestica, the "Jalapeño" chip was developed in just nine months, from the initial design phase to the release of the final version intended for manufacturing.

This chip is intended to be OpenAI's first AI accelerator within a multi-generational computing platform. The three companies are collaborating to make AI faster, more reliable, and more accessible. Thus, this processor represents a significant advancement for OpenAI and its vision for the future of LLM inference.

OpenAI promises that its chip will rely on a robust ecosystem developed by its two partners. Responsibilities will include:

• Industrializing the platform
• Integrating the "Jalapeño" chips into racks
• Networking the chips
• Establishing scalable production systems

A Chip Designed to Be Adaptable and Multi-Generational

The "Jalapeño" chip is designed to be highly adaptable and capable of supporting all major language models. The initial samples produced by Broadcom are already running machine learning workloads, such as GPT-5.3-Codex-Spark, with performance meeting targeted objectives.

However, OpenAI has yet to disclose concrete technical details about this new chip. Specifications remain unknown, but a photo released by the company on X shows the processor with eight HBM slots, positioned on either side of a Compute die.

Expected for deployment by the end of 2026, "Jalapeño" is set to accompany OpenAI for several years, integrated into multiple generations of platforms. The design of this chip is part of a broader industry effort to develop "in-house" AI chips. Google already has custom solutions, while Anthropic is also exploring the creation of bespoke chips.