MEXT's predictive memory software enables transparent auto-tiering of memory pages.
Using AI to make memory intelligent
MEXT predicts an application's near-term memory needs and dynamically allocating memory blocks to optimally trade off cost and performance.
The thesis rode a powerful structural tailwind — the AI datacentre build-out has created massive DRAM shortages, and MEXT offered what DN believed was the only generalised solution for reducing DRAM demand without sacrificing compute, with management estimating customers could save over 37% of memory costs.
We backed a highly experienced technical team led by Gary Smerdon, including one of VMware's principal engineers, the former CPO of Fusion-io (sold to SanDisk for $1.3 billion), and a Senior Vice President of SAP Labs — founders who had built and exited in exactly this layer of the stack before.
Founder testimonial
“DN played a critical role in the development and success of MEXT. As an early investor, they provided important guidance which helped shaped our strategy and direction. The additional support they provided in leading our last round played a critical role in setting the stage for MEXT to be acquired by AMD.”
Gary Smerdon
CEO & Founder
