Custom AI inference accelerator silicon from Positron
Asimov is Positron AI's purpose-built inference accelerator silicon, designed around memory bandwidth and capacity for Transformer decode workloads. It underpins Titan systems and is described on positron.ai/asimov.