GTC 2025 Spring Telecom Special Address

Category: Telecom Special Address · Year: 2025 · ▶ Watch

Speakers: Ronnie Vasishta, SVP, Telecoms

Switch language → 中文

Segments (19)

  • 00:00 · Introduction
    • Opening video highlighting the transformation of telecom networks into AI-powered grids.
  • 01:45 · Global Telco AI Adoption
    • Ronnie Vasishta discusses how telcos worldwide are embracing AI to drive business impact.
  • 03:40 · AI Business Impact for Telcos
    • Overview of how AI improves experience, generates new revenues, and maximizes network ROI.
  • 05:10 · Agentic AI for Network Operations
    • Introduction to Agentic AI platforms and Large Telco Models for automating network operations.
  • 07:30 · SoftBank Large Telco Model (LTM)
    • Announcement and demonstration of SoftBank’s LTM optimizing network traffic for a stadium event.
  • 10:05 · Amdocs smAIz Agents Demo
    • Demonstration of an Amdocs network agent simulating traffic offloading to resolve latency issues.
  • 11:10 · AI Factories and Sovereign AI
    • Discussion on telcos building AI factories to support national sovereign AI agendas.
  • 14:05 · Indosat Sahabat-AI Demo
    • Showcase of Indosat’s local language LLM powering voice-activated services in Indonesia.
  • 16:10 · TELUS Sovereign AI Factory
    • Announcement of TELUS launching Canada’s leading Sovereign AI Factory.
  • 17:20 · Scaling Laws and the AI Grid
    • Explanation of AI scaling laws and the opportunity for telcos to build a distributed AI Grid.
  • 19:05 · Distributed AI Inferencing
    • Examples of AI applications requiring distributed inferencing, such as video search and robotaxis.
  • 21:55 · Latency Requirements for AI
    • Analysis of how different AI applications place deterministic latency requirements on the network.
  • 24:30 · The Shift to AI-RAN
    • Why traditional wireless networks must evolve into multi-purpose AI-RAN architectures.
  • 28:20 · AI-RAN Ecosystem Momentum
    • Highlighting the rapid growth of the AI-RAN Alliance and industry partnerships.
  • 29:55 · AI-Native Wireless Performance Gains
    • Showcasing exponential throughput gains achieved with AI-native wireless networks.
  • 31:15 · 6G Research Partnerships
    • Announcing collaborations to develop an AI-native wireless platform for 6G.
  • 33:05 · New Aerial and Sionna Tools
    • Introduction of new research and commercial development tools for AI-RAN and 6G.
  • 34:50 · Aerial Omniverse Digital Twin Demo
    • Demonstration of simulating RF propagation and network throughput in a 3D digital twin.
  • 37:25 · Conclusion
    • Final thoughts on the Telco AI Renaissance and the call to action for the industry.

Product Announcements (5)

  • [07:30] Large Telco Model (LTM) by SoftBank
    • A domain-specific large language model for telecom network operations.
    • specs: 70B Parameters, trained on NVIDIA DGX SuperPOD, 4TB data corpus covering 200K cells.
    • availability: N/A
  • [16:10] TELUS Sovereign AI Factory
    • Canada’s leading Sovereign AI Factory for businesses and researchers.
    • specs: Powered by NVIDIA Hopper and Blackwell Superchips, 99% renewable energy.
    • availability: First cluster available in Quebec, starting summer 2025.
  • [31:15] 6G AI-Native Wireless Networks Initiative
    • A collaborative effort to research and develop a 6G wireless platform.
    • specs: Partnership with T-Mobile, Cisco, MITRE, ORAN, and Booz Allen using NVIDIA AI Aerial.
    • availability: N/A
  • [33:05] Sionna 1.0 & Sionna Research Kit
    • Open-source library and setup kit for 6G physical layer research.
    • specs: Link-level, ray tracing, and system-level simulation capabilities.
    • availability: Available now via pip install.
  • [33:35] Aerial Omniverse Digital Twin Everywhere & Commercial Test Bed
    • Simulation and testing tools for AI-native wireless algorithms.
    • specs: Available on cloud, on-prem, on laptop; Test bed available on NVIDIA MGX.
    • availability: N/A

Specific Numbers (10)

Timestamp Metric Value Context
07:30 Parameters 70B Size of the SoftBank Large Telco Model (LTM).
07:30 Data Corpus 4TB Amount of data used to train the SoftBank LTM.
13:00 Public Announcements 15 Number of telcos that have publicly announced building AI Factories.
25:45 Monetization 5X Potential revenue multiplier per dollar of Capex over 5 years with AI-RAN.
25:45 Capacity Utilization Up to 3X Improvement in capacity utilization with AI-RAN.
25:45 Energy Efficiency 40% Less Reduction in energy consumption with AI-RAN.
28:20 Alliance Members 75 Number of members in the AI-RAN Alliance within 1 year.
29:55 Throughput Gain 70% Gain achieved using a Learned Profiless Air Interface.
29:55 UL Throughput Gain 50% Gain in low SINR conditions using AI for UL Channel Interpolation.
32:00 Downloads 150K Number of downloads for the Sionna research tool.

Benchmark Claims (4)

  • [25:45] AI-RAN Capacity Utilization: Up to 3X better
    • vs: Traditional RAN systems
    • gain: Significant improvement in network capacity utilization.
  • [25:45] AI-RAN Energy Efficiency: 40% Less
    • vs: Traditional RAN systems
    • gain: Major reduction in energy consumption.
  • [29:55] Learned Profiless Air Interface Throughput: 70% gain
    • vs: Traditional Air Interface
    • gain: 70% increase in throughput.
  • [29:55] AI for UL Channel Interpolation: 50% gain
    • vs: Traditional methods in low SINR
    • gain: 50% increase in uplink throughput.

Customer Stories (4)

  • [07:30] SoftBank
    • Developed a Large Telco Model (LTM) to optimize network configurations.
    • outcome: Successfully simulated and optimized network traffic handling for a major stadium event.
  • [10:05] Amdocs
    • Built smAIz Agents for automated network operations.
    • outcome: Created a digital human agent capable of diagnosing latency issues and simulating traffic offloading.
  • [14:05] Indosat
    • Deployed Sahabat-AI, a local language LLM for Indonesia.
    • outcome: Enabled voice-activated services for ride-hailing and food delivery in Bahasa Indonesia.
  • [16:10] TELUS
    • Launched Canada’s leading Sovereign AI Factory.
    • outcome: Provided sustainable, high-performance AI infrastructure for Canadian businesses and researchers.

Key Technologies (7)

  • Agentic AI: Enables AI agents to autonomously reason, plan, and execute complex network operation tasks.
  • Large Telco Model (LTM): A domain-specific LLM trained on telecom data to assist with network configuration and management.
  • AI Factories: Dedicated, high-performance data centers designed specifically for training and running AI models.
  • AI Grid: A distributed network architecture that allows telcos to provide AI inferencing capabilities at the edge.
  • AI-RAN: A unified architecture that runs both Radio Access Network (RAN) and AI workloads on the same computing infrastructure.
  • Sionna: An open-source Python library for simulating and researching 6G physical layer communication systems.
  • Aerial Omniverse Digital Twin: A physically accurate 3D simulation environment for testing and optimizing wireless network deployments.

Demos Shown (4)

  • [07:30] SoftBank LTM simulating and optimizing network traffic for a baseball game at a stadium.
    • True
  • [10:05] Amdocs smAIz Network Agent diagnosing a latency issue and simulating traffic offloading.
    • True
  • [14:50] Indosat Sahabat-AI voice assistant ordering a ride and food using Bahasa Indonesia.
    • True
  • [34:50] Aerial Omniverse Digital Twin simulating RF propagation and throughput optimization in a 3D city model.
    • True

Predictions / Commitments (4)

  • [16:10, Summer 2025] TELUS Sovereign AI Factory first cluster will be available in Quebec starting summer 2025.
  • [18:15, Future] The AI Grid will democratize AI at a national scale, making telcos key distributors of AI compute.
  • [24:30, Future] Future AI-RAN architectures will be multi-purpose, unified, software-defined, and heterogeneous.
  • [31:15, 6G Era] NVIDIA and partners will research and develop an AI-native, highly spectrally efficient 6G wireless platform.

Companies Mentioned (6)

SoftBank · Amdocs · Indosat · TELUS · T-Mobile, SoftBank, IOH, Ericsson, Nokia, Samsung, Fujitsu · T-Mobile, Cisco, MITRE, ORAN, Booz Allen

Notable Quotes (3)

The Telco AI Renaissance is here. — Ronnie Vasishta @ 01:45

The AI Grid is the biggest telco opportunity. — Ronnie Vasishta @ 18:15

Traditional wireless networks are not designed for AI. A new architecture is needed to serve the intelligence era. — Ronnie Vasishta @ 24:30

Key Topics

Telco AI · Agentic AI · Large Telco Models (LTM) · Sovereign AI · AI Factories · AI Grid · Distributed AI Inferencing · AI-RAN · 6G Research · Digital Twins · Omniverse · Telecommunications

Takeaways

  • Telcos are rapidly adopting AI to improve operations, build AI factories, and modernize wireless networks.
  • Agentic AI and Large Telco Models (LTMs) are enabling autonomous, highly efficient network operations.
  • Telcos are uniquely positioned to build Sovereign AI Factories to serve national and enterprise AI demands.
  • The ‘AI Grid’ represents a massive opportunity for telcos to distribute AI compute capabilities at the edge.
  • Traditional RAN architectures are insufficient; the industry is shifting to unified, software-defined AI-RAN.
  • AI-RAN offers significant ROI through better capacity utilization, energy savings, and new monetization avenues.
  • NVIDIA provides a comprehensive suite of tools, including Sionna and Aerial Omniverse, to accelerate 6G and AI-RAN development.