THE CRUNCH
ADATA TRUSTA is showcasing its AI Scaler Extended Memory solution at AI Infra Summit 2026, aiming to help enterprises run larger AI models on systems with limited GPU memory. The hardware-software architecture integrates GPU, DRAM and SSD resources to optimise utilisation and can reduce deployment costs by more than 50% for AI inference and fine-tuning. The solution supports a range of open-source models and AI agent
The TRUSTA AI Scaler Extended Memory solution uses a hybrid architecture to break through GPU memory limits. It is powered by the AI Scaler Toolkit, which supports mainstream model families including Llama, Qwen and Mistral, and integrates with AI agents such as OpenClaw and NemoClaw. The approach was recognised with the COMPUTEX 2026 Best Choice Award.
Alongside the memory solution, ADATA TRUSTA is presenting the TD7P51 ECO PCIe Gen 5 enterprise SSD and DDR5 R-DIMM. The SSD offers capacities up to 15.36 TB across U.2, E1.S and E3.S form factors and features Flexible Data Placement technology. The R-DIMM delivers speeds up to 6,400 MT/s and capacities from 32 GB to 128 GB, providing high-bandwidth memory for data centre and HPC workloads.