ADATA cuts AI deployment costs 50% with new TRUSTA AI Scaler

The launch at Computex 2026 coincides with a broader industry push toward a 'Hybrid AI' architecture that balances cloud and local processing. Per Digitimes (May 2026), other memory leaders like Phison and Samsung have also joined NVIDIA's ecosystem to optimize storage performance for localized inference. This shift is driven by the scaling demands of large language models (LLMs) which have outpaced the growth of on-chip HBM (High Bandwidth Memory). By using ADATA's toolkit, developers can now run 70B+ parameter models on single-GPU workstations that previously required multi-node clusters.

Furthermore, NVIDIA recently expanded its own agentic AI footprint with the NemoClaw platform. Per GlobeNewsWire (June 2026), NemoClaw offers a secure runtime environment for autonomous agents, which ADATA demonstrated on its Jetson Thor-powered robotics. This ecosystem integration between NVIDIA'S software stack and ADATA's high-speed Gen5 storage emphasizes a focus on 'Physical AI'—where robots and edge devices perform real-time contextual analysis.

On the consumer side, the 25th-anniversary product refresh signals a new focus on 8K raw video workflows and high-speed portability. Per TechPowerUp (June 2026), ADATA’s CUDIMM memory modules are now targeting speeds exceeding 10,000 MT/s, catering to the rising hardware requirements of locally-hosted AI assistants. The inclusion of NFC-encrypted storage in the Urban TapSafe series highlights a secondary industry trend toward physical security layers for sensitive creative assets, moving away from purely cloud-dependent DRM and password systems.