On October 28, Phison Electronics and RedData, an RPI-CS, Inc. division, announced the availability of Phison’s aiDAPTIV+ GPU memory extension technology integrated into RedData’s secure solutions for U.S. federal classified programs.
Following the recent release of America’s AI Action Plan, many federal agencies and system integrators are faced with implementing AI infrastructure using secure and future-proof methodologies. The collaboration between Phison and RedData addresses these critical requirements for affordable, secure, on-premises AI infrastructure within government use cases, the intelligence community and national labs. By leveraging Phison’s aiDAPTIV+ AI solution, federal agencies can affordably accelerate inference and scale large language model (LLM) training while maintaining compliance with classified environment requirements.
Federal agencies face unique challenges deploying AI due to security, compliance, and the prohibitive cost of scaling GPU card-only systems. By combining Phison’s technology and RedData’s AI platforms, this solution delivers a private, cost-effective path for agencies to operationalize AI in mission-critical programs.
“aiDAPTIV+ was engineered to unlock the full potential of GPUs by using NAND flash storage to break through memory constraints,” said Michael Wu, general manager and president, Phison US. “In partnership with RedData, we are bringing this capability directly to U.S. federal customers who require secure, classified AI environments.”
“Our government customers are seeking practical, deployable solutions to enable AI in their most sensitive missions,” said Bill Downer, vice president of sales, RedData. “By integrating Phison’s aiDAPTIV+ into RedData’s secure platforms, we’re empowering agencies to train and deploy larger AI models while meeting stringent classified requirements.”
Source: Phison Electronics
Stay in the know with breaking news from across the IC and IC contracting landscape by becoming a paid subscriber to IC News. Your support makes our work possible.
By 
