Next Generation Memory

More Memory.
Less Power.

IronLattice builds a drop-in replacement for DRAM and HBM. It's fast enough to keep up with AI compute and permanent enough to never forget, manufacturable in fabs that already exist today.

Learn how we did it.

Contact Us
IronLattice wafer
Partners & Collaborators
02

The memory AI has been waiting for.

IronLattice is developing next generation nonvolatile memory architecture for AI and high performance computing. It combines the speed of DRAM with the persistence of NAND, fab friendly by design. That means it runs inside today's standard chip manufacturing (CMOS), no exotic materials, no retooling.

03

Every Computer Has Had to Choose

Built on a Compromise

Computing systems run on two kinds of memory. Fast memory keeps the processor moving but loses everything when power drops. Persistent memory holds data for years but cannot keep up with the processor.

Every handoff between the two costs time, energy, and complexity, and that cost scales as AI does.

The Race Today

IronLattice performance on AI tasks.

DRAMToday's fast tierQuick, then forgets
Important AI Compute Task
Refresh Cycle
NANDToday's storage tierPersistent and slow
IronLatticeFast and remembers
Same Task, Same Prompt
Today's memoryDRAM plus NAND
Reading from NAND
IronLatticeDRAM speed, NAND persistence
Working
04

Made in America

Designed for domestic manufacturing with US sourced materials. No rare earth or conflict minerals.