Phison Electronics has announced expanded capabilities for its aiDAPTIV+ technology, bringing powerful AI processing to notebooks, desktops, and mini-PCs. By combining high-capacity, flash-based memory with integrated GPU architectures, aiDAPTIV+ enables large-model AI training and inference on affordable, mainstream devices.
Expanding GPU Memory for Large-Scale AI Models
The upgraded aiDAPTIV+ technology addresses the growing demand for on-device AI performance by using NAND flash memory to extend GPU capacity and overcome traditional VRAM limitations. This allows developers and users to train and run large-scale models directly on PCs, reducing the need for expensive data-center hardware.
Phison’s collaboration with partners such as Acer demonstrates how aiDAPTIV+ can unlock large language model (LLM) training using significantly less DRAM. For instance, Acer’s testing with the gpt-oss-120b model showed high performance with only 32GB of memory, dramatically improving efficiency for real-time AI applications.
Michael Wu, President and GM of Phison US, noted, “By expanding GPU memory with high-capacity, flash-based architecture in aiDAPTIV+, we offer everyone, from consumers and SMBs to large enterprises, the ability to train and run large-scale models on affordable hardware. In effect, we are turning everyday devices into supercomputers.”
aiDAPTIV+ enhances inference speed by reusing stored tokens, reducing recalculation and power consumption. According to Phison’s data, response times improve up to 10x faster, with a notable decrease in latency.
The platform also allows Mixture of Experts (MoE) models to run efficiently on smaller systems, reducing DRAM use from 96GB to 32GB while maintaining high performance.
At CES 2026, Phison is showing its aiDAPTIV+ technology running on devices from Acer, MSI, ASUS, Corsair, and NVIDIA. The company says this technology helps regular laptops and desktop computers handle demanding AI tasks faster and more efficiently, turning everyday PCs into more powerful and affordable machines for users and businesses alike.
