Technology

Add as a preferred source on Google

Huang Says Telecom Infrastructure Could Support Distributed AI Networks

He also estimated that a one-gigawatt AI data center would cost about $50 billion, including its site, power, buildings, equipment, storage and cooling.

Updated 1 min read
Attendees watch a presentation during a technology conference in Berlin, Germany. / TokenPost.ai (mono)
Attendees watch a presentation during a technology conference in Berlin, Germany. / TokenPost.ai (mono)

Telecom operators may be able to supply computing capacity for artificial intelligence through their base stations and data centers, NVIDIA CEO Jensen Huang said in Berlin, outlining a possible distributed network for AI services.

At GTC on Wednesday, Oct. 21, Huang set out a five-layer view of the AI economy: applications, models, AI factories, computing and energy. A presentation at the event placed NVIDIA in computing, linked Siemens Energy and Schneider Electric with energy, and listed Nscale and Firebird among AI factory builders. Examples also included Mistral, Legora and DeepL as makers of AI models and tools.

Huang put the cost of a data center with one gigawatt of capacity at about $50 billion. He said the estimate covered the site and electricity supply, construction, computing and storage equipment, and cooling.

Related articles

View all