d-Matrix Raptor stacks AI accelerator on custom DRAM

d-Matrix presented Raptor, the first 3D DRAM accelerator for generative inference, at Hot Chips 2026. A TSMC 4nm compute die bonds face-to-face at a 36-micron pitch on custom DRAM, delivering 100 TB/s from 32GB per card. CTO Sudeep Bhoja cited 0.37 pJ/bit interface energy.