SiMa.ai has launched a PCIe HHHL Card to support large language models (LLMs) and multimodal AI in industrial PCs. Developed in partnership with Advantech, the Modalix PCIe HHHL Card uses a standard half-height, half-length form factor for integration into existing industrial infrastructure.

The company is positioning the card as a tool to break down integration barriers for physical AI across the robotics and automotive manufacturing sectors.

The Modalix PCIe HHHL card is designed for real-time vision reasoning and AI workloads while consuming less than 10 watts of power. It features direct GMSL camera interfaces and Ethernet connectivity, allowing the entire AI application to run on the Modalix machine learning system-on-chip. This architecture eliminates host CPU overhead and simplifies the addition of AI capabilities to legacy systems.

“SiMa.ai’s Modalix PCIe card redefines what is possible for industrial AI inference at the edge,” said Brian Wilson, AVP of Sales at Advantech. He noted that the card operates within standard PCIe slot power budgets without requiring auxiliary power connectors or active cooling.

The Modalix card supports up to 16 video channels, making it suitable for manufacturing, transportation, and smart vision applications. It is available in both commercial and industrial temperature variants with memory options of 8GB or 16GB. SiMa.ai stated that the card provides a scalable path for industrial operators to deploy complex AI models in power-constrained environments.