ADATA Technology has launched the TRUSTA AI Scaler Extended Memory Solution, a new software-hardware platform designed to reduce enterprise AI deployment costs and address GPU memory limitations in on-premises AI infrastructure.
The launch marks an expansion of ADATA’s enterprise storage business through its TRUSTA brand, which focuses on AI, cloud, and data center storage solutions.
According to the company, the new platform introduces the AI Scaler Toolkit, a software layer designed to extend AI workloads beyond GPU-only architectures by utilizing system DRAM and high-speed SSD storage alongside GPU memory.
The company said the approach allows enterprises to deploy and fine-tune large language models more efficiently while reducing reliance on expensive high-end GPU configurations.
TRUSTA stated that the solution can lower AI deployment costs by more than 50% in certain inference and model fine-tuning scenarios compared with architectures that rely entirely on GPU memory.
Industry analysts note that growing enterprise adoption of AI is driving demand for scalable and cost-efficient infrastructure, particularly as organizations increasingly move AI workloads from public cloud environments to on-premises and edge deployments.
Concerns around data privacy, compliance, latency, and infrastructure costs are also pushing enterprises to build localized AI systems rather than relying solely on cloud-based services.
According to TRUSTA, the AI Scaler Toolkit dynamically allocates workloads across GPU memory, DRAM, and SSDs to improve resource utilization and scalability. In tested deployment scenarios, workloads that traditionally required multiple GPUs were able to operate using a single GPU combined with expanded system memory.
The company said the platform is designed as a free and open-source toolkit that is not restricted to specific hardware configurations. The solution currently supports several mainstream AI model families including Llama, Qwen, Mistral, Mixtral, GPT-OSS, DeepSeek, Phi, and Gemma.
TRUSTA added that the toolkit also supports emerging AI agent frameworks such as OpenClaw, NemoClaw, and Hermes Agentic, allowing enterprises and developers to integrate the platform into broader agentic AI workflows.
The solution recently received the COMPUTEX 2026 Best Choice Award in the AI category. According to the company, the recognition highlights innovation in software-hardware integration and enterprise AI infrastructure design.
Alongside the AI Scaler launch, TRUSTA also introduced its TD7P51 ECO PCIe Gen5 enterprise SSD at COMPUTEX. The SSD supports capacities of up to 15.36TB and is available in multiple enterprise form factors including U.2, E1.S, and E3.S.
The company said the SSD incorporates Flexible Data Placement (FDP) technology designed to improve data reliability, storage efficiency, and workload stability through intelligent data management.
TRUSTA products have also been validated on several enterprise server platforms globally, according to the company.
