Qualcomm Expands Into AI Inference Hardware

Qualcomm disclosed plans for two new AI processors, the AI200 and AI250, designed specifically for deploying artificial‑intelligence models rather than training them. Both chips leverage the company’s Hexagon neural processing units, which currently power AI features in Qualcomm’s mobile devices, laptops, and telecommunications equipment.

AI200: Immediate Offering with High Memory Capacity

The AI200 is scheduled for release next year. It features 768 GB of RAM and is optimized for AI inference tasks. Qualcomm’s design allows the AI200 to be mounted in a rack configuration where up to 72 chips can operate as a single computing system, a capability comparable to the multi‑chip arrangements used by Nvidia and AMD for GPU workloads.

AI250: Future‑Focused Efficiency Upgrade

The AI250 is slated for a 2027 launch and is described as delivering a "generational leap in efficiency" that will enable significantly lower power consumption compared with current solutions. While specific performance metrics were not disclosed, Qualcomm positions the AI250 as a next‑generation option for data‑center‑scale AI inference.

Strategic Partnership with Humain

Humain, an AI company owned by Saudi Arabia’s Public Investment Fund, announced plans to adopt both the AI200 and AI250 chips for building AI datacenters across the kingdom. This partnership underscores Qualcomm’s intent to penetrate the enterprise and cloud AI market and to provide an alternative to Nvidia’s dominant offerings.

Market Implications

By extending its Hexagon technology from mobile devices into high‑performance data‑center hardware, Qualcomm aims to diversify its product portfolio and challenge Nvidia’s market leadership in AI acceleration. The collaboration with Humain also highlights a growing interest from sovereign wealth funds in establishing domestic AI infrastructure.

Usado: News Factory APP - descubrimiento de noticias y automatización - ChatGPT para Empresas