ADATA TRUSTA Unveils AI Scaler Extended Memory for Cost-Effective AI Deployment

ADATA, through its TRUSTA division, has unveiled a new infrastructure solution for artificial intelligence called the AI Scaler Extended Memory Solution. The technology tackles one of the biggest hurdles in modern AI systems: the limited video memory found on graphics accelerators.

Running large language models today requires expensive GPUs with massive VRAM, making corporate AI deployment prohibitively costly. TRUSTA takes a different approach by leveraging not just GPU memory, but also system RAM and fast SSDs to store and process model data.

At the heart of the platform is the AI Scaler Toolkit, an open-source software and hardware package that balances workloads across GPU, DRAM, and SSDs. This allows AI models to run more flexibly and at a lower cost. According to the company, in some scenarios this cuts AI deployment expenses by more than 50 percent.

ADATA says the technology can execute tasks that previously required multiple graphics processors on a single GPU, supported by expanded system memory. This is particularly important for businesses aiming to set up local AI systems without building costly server clusters.

The platform supports popular models such as Llama, Qwen, Mixtral, Mistral, DeepSeek, Gemma, and Phi, and is compatible with agent-based AI systems like OpenClaw and Hermes Agentic. The developers stress that the solution is not tied to any specific hardware and can adapt to various server configurations.

Alongside this, TRUSTA also introduced the TD7P51 ECO PCIe Gen5 enterprise SSD, available in capacities up to 15.36 TB. The drive supports multiple form factors and uses intelligent data placement technologies to enhance reliability under AI workloads.

The company believes that such solutions will help accelerate the shift from cloud-based AI services to on-premise systems, where data control, security, and cost savings are key priorities.