Ntt Inc · TSE:9432
Ntt — Corporate Update (October 2025)
18 slides from Ntt's corporate update, Oct 6, 2025
Slide 1: Computing Infrastructure for the AI Era
Computing Infrastructure for the AI Era Sachiko Oonishi Executive Member of the Board, Executive Vice President, CCXO and Co-CAIO Head of Research and Development Market Strategy NTT, Inc. Oct. 6th, 2025 © NTT, Inc. 2025
Slide 2: The Rapid Surge in Power Consumption Due to the Wider Adoption of AI
The Rapid Surge in Power Consumption Due to the Wider Adoption of AI • GPU servers consume 5.9 times the power of traditional servers. • GPU installations are growing 1.8x annually, driving continuous power consumption increases. Power consumption of a server (*1) Scaling of AI computers(*2) xAI 1,3
Slide 3: The Rapid Adoption of AI Is Fueling a Power Crisis
The Rapid Adoption of AI Is Fueling a Power Crisis • At this rate, data center power will surpass Tokyo's 2023 annual consumption. Forecast of data center power consumption ［100M kWh］ As is Data center Network Modest 1000000 Optimistic 100000 120000 88000 10000 5000 6000 900 1000 1000 1100 1000 Toky
Slide 4: As AI Use Expands, User Costs Are Also Surging
As AI Use Expands, User Costs Are Also Surging • As AI adoption grows, surging utilization costs are fueling concerns about declining ROI. • The infrastructure domain will account for over half the AI market. Enterprise AI adoption costs (*１) Forecast of (Average Annual LLM Budget per Company) AI ma
Slide 5: Requirements for Infrastructure in the AI Era
Requirements for Infrastructure in the AI Era • Efficient operation and reduced energy consumption are essential for AI computing infrastructure. • IOWN is a crucial component for balancing AI benefits with efficient operation and reduced power consumption. Efficient Infrastructure CPU Memory CPU Me
Slide 6: Wider AI Adoption is Reshaping Compute Utilization
Wider AI Adoption is Reshaping Compute Utilization • As AI use expands across business processes, fluctuating compute utilization drives up power and costs. Previously: Limited AI applications led to The expansion of AI and task diversity is causing fluctuating high GPU utilization for specific task
Slide 7: The Reality of Compute Infrastructure in the AI Era
The Reality of Compute Infrastructure in the AI Era • Growing infrastructure demand, led by AI, is maxing out power supply in the Tokyo Metropolitan Area. • Data centers and GPUs are increasingly being distributed to regions with power surpluses. GPU GPU GPU GPU © NTT, Inc. 2025 6
Slide 8: Infrastructure Optimization Initiatives at Leading Companies
Infrastructure Optimization Initiatives at Leading Companies • Leading companies are starting to efficiently operate infrastructure by visualizing resource status (e.g., GPU usage and power consumption) and prioritizing AI workloads in resource-rich locations. Site B Visualization of Resource Usage
Slide 9: Infrastructure Requirements in the AI Era: 1. Efficient Infrastructure Operations
Infrastructure Requirements in the AI Era: 1. Efficient Infrastructure Operations • High-speed APN links to connect multi-tiered GPUs and distributed GPUs. • Flexible allocation of compute resources, such as GPUs, based on AI application and utilization. • Dynamically assign AI tasks to sites with s
Slide 10: As AI Utilization and Data Volume Increase, Electrical Processing Is Reaching Its Limit
As AI Utilization and Data Volume Increase, Electrical Processing Is Reaching Its Limit • Increased AI processing → More GPUs → Greater intra-computer communication • Increased power consumption and heat generation present a limit for electrical communication → Transition to optical communication. b
Slide 11: Infrastructure Requirements in the AI Era: 2. Reducing Power Consumption
Infrastructure Requirements in the AI Era: 2. Reducing Power Consumption • By implementing Photonics-Electronics Convergence devices (PEC-2) in network switches, we can achieve both increased communication capacity and reduced power consumption. Inter-cluster interconnect Reduce power Inter-cluster
Slide 12: AI Adoption Is Progressing Steadily
AI Adoption Is Progressing Steadily Number of （Number of Orders） Generative AI Sales Order Status Domestic + Overseas（Number of Inquiries) Orders 2,000 3,000 Number of Domestic Orders 1,827 ※1 Number of Overseas Orders Number of Domestic Inquiries Number of Domestic Inquiries 1,818 1,000 1,326 1,306
Slide 13: NTT's GPU as a Service: Delivering Compute Resources When and Where Needed
NTT’s GPU as a Service: Delivering Compute Resources When and Where Needed, Based on AI Utilization Custom-built based on A Dept. the required number of B Dept. Edge C Dept. GPUs Enterprise A Enterprise B Enterprise C Enterprise D Devices Delivering optimized GPU setups as a managed service Single-T
Slide 14: Jointly Developing and Operating a "Mobility AI Platform" with Toyota
Jointly Developing and Operating a "Mobility AI Platform" with Toyota to Realize a Zero-Traffic-Accident Society Distributed Computing Platform Private Cloud Services On-Prem Data Centers © NTT, Inc. 2025 Public Cloud Services 13
Slide 15: Computing Infrastructure for the AI Era
Computing Infrastructure for the AI Era • Efficiently managing compute resources to boost GPU utilization and reduce overall power consumption. • Introducing Photonics-Electronics Convergence devices enables further reduced power consumption by running the same infrastructure with optimized GPUs and
Slide 16: Innovating a Sustainable Future for People and Planet
Innovating a Sustainable Future for People and Planet
Slide 17: NTT
©) NTT
Slide 18: This document is a translation of the Japanese original
This document is a translation of the Japanese original. The Japanese original is authoritative. The forward-looking statements and projected figures concerning the future performance of NTT and its subsidiaries and affiliates contained or referred to herein are based on a series of assumptions, pr