d-Matrix Introduces JetStream I/O Accelerators for Large-Scale AI Inference with PCIe Gen5 and 400Gbps Bandwidth

d-Matrix has announced the addition of JetStream I/O accelerators to its AI product portfolio, designed to enable data center-scale AI inference for models exceeding 100 billion parameters. The JetStream I/O card, a Transparent NIC and streaming solution, is optimized for d-Matrix’s Corsair accelerators and Aviator software, targeting ultra-low latency deployment across multiple users as AI services shift focus from model training to inference, particularly for agentic AI, reasoning, and multi-modal interactive content.

JetStream is packaged in an industry-standard PCIe Gen5 full-height format, delivering up to 400Gbps bandwidth and compatibility with off-the-shelf Ethernet switches, allowing deployment in existing data centers without infrastructure replacement. When integrated with Corsair accelerators and Aviator software, JetStream supports up to 10x faster processing, 3x better cost-performance, and 3x higher energy efficiency compared to GPU-based solutions for AI inference workloads.

Sid Sheth, d-Matrix co-founder and CEO, stated that JetStream addresses the growing demand for multimodal AI and user expectations for rapid interactivity, complementing the Corsair compute platform to provide a scalable solution. The combination of JetStream, Corsair, and Aviator positions d-Matrix among the few AI infrastructure providers offering a platform that int...