jueves, 25 de junio de 2026

OpenAI Broadcoms Jalapeno LLM Inference Chip Unveiled

Big news in the AI hardware world! OpenAI, the powerhouse behind ChatGPT, has teamed up with semiconductor giant Broadcom to unveil a brand-new chip specifically engineered for large language model (LLM) inference. Codenamed 'Jalapeño', this Application-Specific Integrated Circuit (ASIC) marks the beginning of a long-term collaboration aimed at revolutionising how LLMs operate within data centres.

OpenAI & Broadcom's Jalapeño: LLM Inference Chip Unveiled

This isn't just another general-purpose chip. Broadcom claims Jalapeño was designed from the ground up, incorporating deep insights from OpenAI's researchers and their future model roadmap. The development process was impressively swift, taking just nine months. The core promise of Jalapeño is to deliver significantly better performance per watt compared to existing inference solutions. While early testing hints at this substantial improvement, OpenAI is still finalising its detailed technical report, which is expected in the coming months. This move aligns with OpenAI's broader strategy of vertical integration, aiming to gain more control over its technology stack and reduce reliance on external suppliers like Nvidia.

The global compute crunch has intensified the interest in custom silicon. For companies like OpenAI, custom chips offer a pathway to unlock greater capacity and efficiency. Broadcom, already a major player in infrastructure silicon, has been actively pivoting towards providing tailored chips for hyperscalers and leading AI model developers during the current AI boom. Both companies are aiming for a significant deployment of Jalapeño chips in data centres by the end of this year, signalling a strong commitment to this new era of AI-specific hardware.

Fuente Original: https://arstechnica.com/gadgets/2026/06/openai-and-broadcom-announce-chip-designed-for-llm-inference-at-scale/

Artículos relacionados de LaRebelión:

Artículo generado mediante LaRebelionBOT

No hay comentarios:

Publicar un comentario