ironlattice has raised $3.0M to build nonvolatile memory it says can close the long-standing gap between speed and persistence. The fabless company is designing an architecture that combines DRAM-level speed with NAND-style retention, aimed squarely at the memory demands of AI compute.
The pitch hinges on manufacturability: ironlattice says its design uses no exotic materials and requires no retooling, so it can run inside standard CMOS chip production. That's a deliberate contrast to approaches that demand new fabrication processes. The company argues AI's appetite for compute infrastructure has made the tradeoff between fast memory and memory that holds data without power increasingly untenable, and that its architecture delivers both.
