IBM released its distributed inference platform on September 28, 2026, stating it is capable of processing large-scale AI workloads with enhanced performance. It is the company's first major hardware update in the AI inference space since 2024.
IBM reported a 30% improvement in inference latency, measured on the MLPerf benchmark. That compares with the 45% latency observed in its previous platform from 2024.
The platform is built on IBM's latest semiconductor architecture and targets enterprise AI applications requiring high throughput and low latency. Availability begins in the fourth quarter of 2026, initially for select enterprise clients.
"Distributed inference enables us to scale AI workloads efficiently across multiple nodes," said Peter Hess, IBM's hardware lead. "This platform is designed to support complex AI tasks without compromising performance."
The announcement follows IBM's recent focus on expanding its AI infrastructure. IBM emphasized the importance of scalable and efficient AI solutions for enterprise growth.
IBM did not specify the exact pricing model for the platform, and raised the question of how enterprises will integrate the new system into existing workflows. The company said it will provide more details in the coming months.
Source: ibm
