KM news
On September 24, 2026, NVIDIA unveiled "Project Mnemosyne," a hardware--accelerated knowledge management (KM) architecture designed for the "Agentic Enterprise." This breakthrough utilizes dedicated "Knowledge Processing Units" (KPUs) to maintain real--time, vector--indexed representations of all organizational data streams at the silicon level. By offloading semantic indexing from general--purpose CPUs, the system reduces the energy consumption of continuous knowledge curation by 90% while enabling sub--millisecond retrieval for autonomous AI agents. NVIDIA confirmed that the first KPU--integrated server racks are scheduled for delivery in the first quarter of 2027.