AMD Acquires Mext to Lower AI Infrastructure Costs

AMD has officially confirmed the acquisition of the promising startup Mext. 

The acquired company's primary specialization is memory allocation optimization in artificial intelligence systems. This deal is designed to help operators of major data centers tackle the inefficient utilization of DRAM modules and bypass the growing component shortages in the market. 

The entire Mext development team will transition to AMD.

How Mext Technology Works

At the core of the startup's developments lies a unique tiered memory management technology that leverages functions such as intelligent transfer and AI prediction.

In the case of intelligent transfer, the system continuously analyzes data access patterns and automatically moves rarely used data from fast and expensive DRAM into NAND flash memory.

AI prediction works as follows: built-in algorithms calculate which data sets the processor might require in the next few milliseconds and proactively retrieve them back into DRAM.

This approach significantly boosts the performance of already deployed hardware without requiring the purchase of additional memory modules, ultimately lowering the total cost of ownership (TCO) for cloud providers.

The Response to Competitors

Integrating Mext's technologies into AMD's portfolio represents a major strategic move.

Similar concepts for intelligent data exchange between DRAM and NAND already exist on the market, such as Phison's aiDAPTIV+. 

However, these competitors are strictly locked into their own proprietary drives, whereas AMD has the opportunity to build a much more universal platform.