Z2RAM: Gelsinger-Backed Memory Aims to Ease the AI Chip Crunch

Nagare Labs, a startup backed by former Intel CEO Pat Gelsinger, is developing a new memory type called Z2RAM that aims to ease the shortage driven by the AI boom. The company says it blends the speed of SRAM, the density of DRAM and the vertically stacked architecture of NAND in a single package.

The semiconductor industry is going through one of its worst crunches since the pandemic. Soaring AI demand for memory has caused shortages and price increases across all segments, smartphones included.

How Z2RAM works

Conventional DRAM stores each bit with one transistor and one capacitor. Z2RAM uses two transistors instead, relying on the gate capacitance of the second one to hold the charge. That removes the need for a separate capacitor, and the data does not have to be rewritten after every access, a known limitation of DRAM.

The design is meant to be stacked vertically, like NAND and other 3D packaging approaches, using oxide semiconductors and three-dimensional hybrid bonding. According to Nagare Labs, this could deliver more than five times the bit density of SRAM at similar bandwidth: an area that holds 64 MB of SRAM could hold 320 MB of Z2RAM. The company also claims much lower power consumption.

Why it matters for AI

Beyond raw compute, a key bottleneck for AI accelerators is how quickly data can be moved and stored. The more fast memory sits close to the processor, the fewer chips are needed to run advanced models.

Z2RAM would not replace conventional memory, but it could reduce the number of accelerators data centers need. That would ease pressure on the supply chain, improving component availability and potentially lowering prices industry-wide.

A long road ahead

The technology is still at an early stage. Nagare Labs engineers have built first prototypes, but commercial production could take a decade. The company still has to solve reliability issues and standardize interfaces. If Z2RAM delivers on its promises, it could become one of the more important hardware innovations of recent years.