Advanced Micro Devices (AMD) acquired MEXT Corp. on June 15, 2026, to integrate predictive memory optimization into its data center offerings [1], [3].

This acquisition addresses the critical memory bottlenecks currently hindering artificial intelligence and large-scale data center workloads. By integrating MEXT's technology, AMD aims to lower the high costs associated with memory while maintaining the performance levels required for modern AI applications [1], [4].

At the center of the deal is MEXT's Predictive Memory Engine. This technology enables a process called memory tiering, which identifies infrequently accessed data and offloads it from the system's Dynamic Random Access Memory (DRAM) to NAND flash storage [1], [2]. This mechanism allows flash storage to appear as DRAM to applications, effectively expanding the available memory pool without requiring a proportional increase in expensive hardware [1], [3].

Data center operators often face a trade-off between the speed of DRAM and the capacity of flash storage. The MEXT technology bridges this gap by predicting which data is needed and moving it between tiers automatically [4]. This optimization is intended to ease the growing memory constraints that occur as AI models become larger and more complex [1], [4].

AMD plans to incorporate this predictive memory optimization directly into its AI stack [4]. This move follows a trend of hardware providers seeking software-defined ways to maximize existing silicon efficiency. By making flash behave like DRAM, AMD can provide a more scalable memory architecture for enterprises managing massive datasets [1], [3].

“AMD acquired MEXT Corp. on June 15, 2026, to integrate predictive memory optimization into its data center offerings.”

The acquisition signals a strategic shift toward software-defined memory management to bypass physical hardware limitations. As AI models grow, the cost and power consumption of DRAM become prohibitive; by using predictive tiering to make cheaper flash storage act like high-speed memory, AMD can offer a more cost-effective scaling path for AI infrastructure compared to relying solely on physical memory expansion.