Enabling cutting edge AI by stacking memory
Models are now running into the trillions of parameters, and the world needs a new way of approaching memory to make it possible with both extreme performance and extreme volume.
The premium ultra-low-latency experiences powered by frontier models that users demand only work if there’s a way to scale it gracefully. Doing it with DRAM alone won’t cut it—but a big pile on top of each other will. See how stacking DRAM delivers an SRAM-level upgrade to inference powered by massive trillion-plus parameter models.
Welcoming Wallaroo to the d-Matrix team
Wallaroo, a leader in AI inference deployment and orchestration software, is joining us to deliver a seamless experience building pipelines augmented with the ultra-fast, extraordinarily efficient performance that Corsair can deliver.
Partnering with Callosum to accelerate heterogeneous inference at scale
Our mission is to accelerate lightning-fast inference with heterogeneous hardware pipelines, and we're partnering with Callosum to do it. d-Matrix is now part of Callosum's new global partner network with AMD, Intel, and others.
Model sizes are exploding—and we have to rethink memory to keep up.
Interested in more from d-Matrix?
Get the latest d-Matrix updates to your inbox. Sign up below:
Breaking down Raptor at Hot Chips 2026
d-Matrix co-founder Sudeep Bhoja walked through our next act: our 3D DRAM enabled Raptor accelerator. See more about our next AI inference chip powered by stacked DRAM.
