Google: AI efficiency gains collide with surging infra demand
Google says the AI era is changing how hyperscale infrastructure is designed and operated, making efficiency as critical as adding new computing capacity. In its 2026 Environmental Report, the company argues that AI growth will depend not only on building more data centers, but also on reducing the energy and infrastructure overhead required to run increasingly power-hungry workloads.
The report underscores the scale of AI’s impact. Google’s electricity consumption increased 37% year over year in 2025 as AI services expanded rapidly. At the same time, the company says it matched 100% of its annual electricity use with renewable energy purchases for the ninth consecutive year while reducing operational emissions by 2%. These figures highlight the growing challenge of balancing AI expansion with sustainability goals.
For AI infrastructure operators, one of the report’s most significant metrics is Google’s fleet-wide average power usage effectiveness (PUE) of 1.09 in 2025. Google says this represents 83% less overhead energy for cooling, power distribution, and other non-IT functions than the industry average PUE of 1.54, cited from the 2025 Uptime Institute survey. A PUE close to 1.0 means a greater share of incoming electricity is used directly by IT equipment rather than facility operations.
Google attributes these gains to improvements across its AI stack, including custom TPU accelerators, data center engineering, and software optimization. According to the company, its seventh-generation Ironwood TPU is nearly 30 times more power efficient than its first Cloud TPU, while the energy required to serve the median Gemini text prompt fell by a factor of 33 over the past year.
The report also highlights the growing role of AI infrastructure in power systems. Google has integrated more than 1 GW of demand-response capacity into utility agreements and signed over 70 clean energy agreements during 2025, reflecting the industry’s increasing focus on ensuring AI data center growth can be supported by both efficient facilities and expanding clean power supplies.
Juan Pedro Tomas
Editor
RCRTech
AI Infrastructure Top Stories
AI control plane: Sean Kinney, again – live from Cisco Live some weeks ago; but the takesways are pertinent: that Cisco is making bigger bets on the network (“more powerful than the node”) in an AI world where comms, security, identity, observability and cost converge into a single operational problem.
Bangkok AI capacity grows: True IDC is expanding its data center footprint in Thailand with a seventh facility in northern Bangkok, designed for AI, hyperscale computing, and liquid cooling as regional digital infrastructure demand continues to grow.
AI Today: What You Need to Know
SoftBank’s U.S. neocloud: SoftBank forms SB Neo to launch a U.S. neocloud business, leveraging its planned 10 GW AI infrastructure platform to deliver large-scale AI training and inference services beginning in FY2027.
Digital Realty deepens Virginia bet: Digital Realty is increasing its ownership of three fully leased Northern Virginia hyperscale data centers through a $3.5 billion transaction, expanding its position in the world’s largest data center market.
New Japan DC platform: DigitalBridge and JEXI have launched Nippon Gateway Infrastructure, a new colocation platform that starts with NEC assets and plans to expand enterprise-focused data center capacity across Japan.
Schneider expands industrial AI: Schneider Electric will acquire industrial AI software company Cognite for $3.1 billion, strengthening AVEVA’s industrial intelligence platform with expanded data contextualization and agentic AI capabilities.
Vietnam builds AI campus: Vietnam’s G-Group will invest $300 million in an AI-focused data center campus near Hanoi, combining Tier III infrastructure, HPC capacity, and cloud services to support national AI ambitions.
Africa’s DC opportunity: Africa’s data center investment market could reach $8.76 billion by 2031, although power availability, grid infrastructure, and regulation remain key challenges despite growing connectivity and digital demand.
AI optimizes DC cooling: NTT Data and Daikin will test an AI-based cooling system that predicts server heat and coordinates cooling equipment, aiming to improve efficiency before commercialization in fiscal 2027.
RCR Events
Industry Resources
Webinar, July 7th: Noise-Figure Measurements with RFmx and PXI VSTs
Webinar, July 16th: NTN in motion — evolving standards, expanding services
Whitepaper: Powering sovereign AI at scale
Whitepaper: Scalable database design for 5G and beyond
Report: Scaling AIOPs from insight to action
Summit Access: GSMA Device Enablement Summit: How operators can fix device-network fragmentation
Whitepaper: Telco AI Enabler: Mediation’s defining role
Report: Securing telecom infrastructure for the quantum era
Report: Scaling optical networks for the AI and hyperscale era