Nvidia overhauls the data center for OpenClaw era

Nvidia unveiled a new data center architecture at its GTC conference to support the demands of multi-agent AI systems and technologies such as OpenClaw. Chief executive Jensen Huang described the shift during a keynote, saying data centers once stored files but now function as factories to generate tokens, the units of data processed The architecture incorporates new chips including the Rubin GPU and Vera CPU, along with the Groq LPU inference chip. It also features doubled NVLink interconnect speeds reaching 260 tera Analyst Jack Gold of J. Gold Associates said the push into AI data centers matters because costs and complexity continue to rise. Sandip Gupta of NTT Data noted that an integrated stack simplifies operations for customers who prefer that approach over mixing components. Nvidia said the changes also target efficiency at the edge and in space while responding to AI agents that require real-time access to contextual data.