Jensen Huang: Future telecom base stations are expected to develop into distributed AI networks, and all telecom operators will provide AI computing power

On October 21, NVIDIA CEO Jensen Huang stated at the GTC conference held in Berlin that in the future, all telecom operators will provide AI computing power, and telecom base stations are expected to develop into distributed AI networks. He compared the AI economy to a five-layer cake composed of energy, computing power, AI factories, models, and applications, and pointed out in his presentation that only NVIDIA's logo appeared on the computing power layer. The other layers included energy companies such as Siemens Energy and Schneider Electric, AI factory builders Nscale and Firebird, as well as European AI model company Mistral, legal AI company Legora, and translation tool DeepL. Jensen Huang also mentioned that the cost of building a 1-gigawatt data center is currently about $50 billion, covering expenses such as land, electricity, construction, computing equipment, storage, and cooling.

At the same time, NVIDIA launched the DSX Pod AI factory to accelerate the deployment of AI infrastructure in Europe; the DSX Pod can shorten deployment time to as fast as 14 days; the European industrial ecosystem is collaborating with NVIDIA DSX to create the most energy-efficient AI factory in the world.

According to reports, its DSX AI factory platform uses closed-loop liquid cooling and dry coolers, reducing the facility's cooling water usage to nearly zero, saving up to 54 million liters of water per day at a scale of 2 gigawatts. The DSX MaxLPS software can help operators reduce power consumption by up to 30% under the same token output.

NVIDIA introduced the OpenShell reference architecture, integrating the Vera Rubin infrastructure, Vera Central Processing Unit (CPU), BlueField-4 Data Processing Unit (DPU), and OpenShell secure runtime software to help enterprises build, run, and manage AI agents at scale, while monitoring and enforcing security constraints on their behavior. The architecture continuously monitors agent behavior through the NVIDIA Sentry security framework running on BlueField-4 and enforces security policies at the hardware level. When an agent attempts to breach the boundaries set by the software, the system can isolate and stop it within milliseconds. NVIDIA claims that the Vera Rubin NVL72 system can achieve up to 30 times the previous throughput per megawatt under agent workloads, and NVIDIA Sentry can increase agent capacity by up to 20%. This architecture supports enterprise environments such as Linux, Kubernetes, Microsoft Windows, and can be scaled to edge devices. Companies like Microsoft, Google Cloud, Dell Technologies, Cisco, OpenAI, and Anthropic are participating in related collaborations.

Updating…….