Which utilities are furthest along with AI?

In the Enterprise AI diffusion report, Dominion Energy, Duke Energy, Exelon, Iberdrola, National Grid, NextEra Energy, PG&E, and Southern Company are at Experimenting. A stage follows from the returns and rollouts a company has publicly reported in the last 18 months, not from its spending or plans.

How do utilities use AI?

Utilities use AI most in outage prediction and restoration, vegetation management, line and plant inspections, grid planning and contact centers. Exelon runs its POSEIDON restoration model at all 6 of its utilities (Dec 2025), National Grid uses satellite imagery and AI to cut tree-related outages in New England and scores every New England contact-centre call with AI (May 2026), Dominion Energy applies visual AI to drone imagery (Mar 2026), Duke Energy uses AWS-managed AI agents in interconnection studies (Sep 2026), and PG&E gives 900 contact center agents Agent Assist and 1,300 Diablo Canyon workers the Neutron document search. Each of these utilities is at Experimenting, because none has reported a return it credits to AI alone since May 2025.

What return have utilities reported from AI?

Utilities have reported no return to their own costs, revenue or headcount that they credit to AI alone since May 2025. Duke Energy says AI-driven optimization had saved customers approximately $500 million as of 2024 (Jul 2026), and NextEra Energy says GridComposer saved more than $20 million in 2026, money that went to customers (Sep 2026). PG&E credits $11.2 million saved from Jan 2025 to Jun 2026 to continuous monitoring that joins sensors with machine learning, Iberdrola credits a 75% cut in inspection response times and operating costs on Brazilian transmission lines to drones and AI together (Feb 2026), and NextEra Energy's about $150 million a year of REWIRE and AI savings is a 2026 to 2032 target, not a result.

Which AI platforms and models do utilities use?

Utilities mostly build AI with a major cloud provider and name few model vendors. NextEra Energy runs REWIRE on Google Cloud with Gemini Enterprise Agent Platform, TimesFM 2.5 and WeatherNext 2 (Dec 2025), Iberdrola uses Anthropic Claude and Amazon Nova through Amazon Bedrock (Aug 2024) alongside Microsoft Copilot, Duke Energy builds on AWS, Southern Company uses Databricks for data and AI (Feb 2026) and Microsoft Copilot at Southern Nuclear, and National Grid builds generative AI on Microsoft Azure (Sep 2024). Exelon runs its own SPARK platform and NVIDIA-based vision models on its drones, and PG&E runs Atomic Canyon's FERMI models on NVIDIA GPUs at Diablo Canyon.

How utilities deploy AI

Utilities deploy AI on the grid, in outage restoration, vegetation management and inspections, and count the results in trucks and trees: NextEra Energy avoided 44,000 truck rolls in a year, and National Grid has nearly 30% fewer tree-related outages in New England. Much of the money goes to customers, such as Duke Energy's approximately $500 million of AI-driven savings. NextEra Energy now sells its grid AI to other utilities through Google Cloud Marketplace and keeps all of the intellectual property.

Verified Nov 5, 2026See methodology

How utilities have deployed AI

Findings

What the utilities profiles show

No utility reports a return to its own costs, revenue or headcount that it credits to AI alone since May 2025; the sector's figures are customer savings, modelled values, targets or mixed credits. Duke Energy's approximately $500 million of AI-driven optimization savings as of 2024 (Jul 2026) and NextEra Energy's more than $20 million GridComposer saved in 2026 (Sep 2026) went to customers, NextEra Energy's roughly $450 million of NPV gains on 2024 projects (Dec 2025) is a modelled value and its about $150 million a year of REWIRE and AI savings is a 2026 to 2032 target, PG&E's $11.2 million from Jan 2025 to Jun 2026 is credited to continuous monitoring with sensors and machine learning, and Iberdrola's 75% cut in inspection response times and operating costs on Brazilian transmission lines (Feb 2026) is credited to drones and AI together.
Utilities report AI results as operating measures with no cost attached: National Grid reports nearly 30% fewer tree-related outages in New England from satellite imagery and AI (May 2026), Exelon POSEIDON restoration estimates up to 3 times more accurate in severe weather (Sep 2026) and drone inspection planning cut from one hour to 30 seconds (Dec 2025), Duke Energy interconnection data preparation cut from about 30 days to hours with agentic AI (Jul 2026), PG&E calls 4% shorter with Agent Assist (May 2026) and Diablo Canyon safety-valve document searches cut from about 180 days to 40 with Neutron (Mar 2026), and NextEra Energy 44,000 FPL truck rolls avoided using AI in 2024 (Dec 2025).
Utilities count AI projects and tool access rather than AI users across the company: Iberdrola counts more than 450 AI projects (Jul 2026) and nearly 500 agents deployed or in development (Aug 2026), NextEra Energy more than 100 strategic AI initiatives under execution (Oct 2026), and Duke Energy more than 50 AI products in production (Aug 2025) and a generative AI tool given to about 10,000 of its 26,441 employees in 2025. The user counts that exist cover one unit: more than 850 active Microsoft Copilot users at Southern Nuclear (Dec 2025), and 900 contact center agents on Agent Assist and 1,300 Diablo Canyon workers on Neutron at PG&E.
Utilities split between building AI platforms with a cloud provider and buying AI inside vendor tools: NextEra Energy runs REWIRE, GridComposer and its Optos products (Sep 2026) on Google Cloud with Gemini Enterprise Agent Platform, TimesFM 2.5 and WeatherNext 2, Iberdrola built a generative AI platform on Amazon Bedrock with Anthropic Claude and Amazon Nova (Aug 2024) and added agents on Amazon Bedrock AgentCore (Feb 2026), Duke Energy builds on AWS with an internal platform called AI Central (Aug 2025), Southern Company standardized its data and AI platform on Databricks (Feb 2026), and Exelon built SPARK in house (Sep 2026). National Grid buys most of its AI inside vendor tools such as AiDASH and NICE CXone, and PG&E buys its generative AI from AWS, Atomic Canyon and PolyAI while building its own machine learning models.
Ownership of AI at utilities is recent or unnamed: Dominion Energy named John Russell Chief AI and Digital Officer in Jul 2026, Alice Delahunty became National Grid's Chief Technology and Innovation Officer on Sep 1, 2026, Southern Company's Technology Organization moves under Chief Legal Officer Sterling Spainhour on Jan 1, 2027, and NextEra Energy's REWIRE sits under Chief Risk Officer Petter Skantze. Duke Energy and Exelon name no executive who owns AI, and Exelon CEO Calvin Butler said in Sep 2026 that he would not discuss how Exelon will deploy AI internally. Dominion Energy's board listed NextEra Energy's data analytics and AI deployment among its reasons to combine the 2 companies (Jul 2026), with a close expected in the second half of 2027.

Stages

Where utilities stand

Experimenting
Dominion Energy, Duke Energy, Exelon, Iberdrola, National Grid, NextEra Energy, PG&E, Southern Company
Experimenting means the company has reported neither a business return it credits to AI nor a company-wide AI rollout with a user count. Pilots, partnerships, plans and spending alone keep a company at Experimenting.

Compare

Utilities side by side

Company	Ownership	Platform	Models	Hosting	Reach	Stage
Dominion Energy	John Russell, VP and Chief AI and Digital Officer (Jul 2026); IT under SVP Wendy Wellener in Dominion Energy Services	Hybrid: Approved generative AI tools, including Microsoft Copilot; misgrid model built in-house	Buzz Solutions PowerAI for inspection imagery (2026); Microsoft Copilot (2024 job posting); others not disclosed	Not disclosed; Copilot is a Microsoft cloud service, a 2026 job ad cites AI in Dominion's private cloud, and the data platform spans cloud and on-premises	Not disclosed; about 15,200 employees (Dec 2025)	Experimenting
Duke Energy	No named AI chief; SVP and CIO Richard Donaldson leads digital strategy with a stated focus on AI; CAO Bonnie Titone oversees IT	Hybrid: Unnamed generative AI tool; AI Central platform; Duke Energy Explorer; AWS-managed agents for grid studies	Not disclosed; Copilot and Genie assistants named (2025)	AWS: primary cloud since 2022 and manager of the grid study agents (2026); Azure for a 2021 methane platform; employee tool hosting not disclosed	About 10,000 employees with a generative AI tool (2025)	Experimenting
Exelon	No named AI owner; IT and cybersecurity under EVP Tim Peterson since Feb 2026	Hybrid: SPARK, POSEIDON and outage models built in-house; drone AI from Deloitte	Not disclosed; in-house predictive models; NVIDIA-based vision models for drones	Not disclosed for SPARK, POSEIDON or the outage and cable models; drone vision models run on NVIDIA Jetson modules aboard the drones; Exelon bought an NVIDIA DGX-1 system before 2023	POSEIDON at all 6 utilities; SPARK user count not disclosed	Experimenting
Iberdrola	Global CIO Sergio Merchán's IT department, with Xabier Muruaga as Global Head of AI & Data	Hybrid: Own generative AI and agent platform on Amazon Bedrock; Microsoft Copilot bought	Anthropic Claude and Amazon Nova via Amazon Bedrock; Microsoft Copilot	AWS (Amazon Bedrock, SageMaker and AgentCore inside a VPC) and Microsoft Azure; no on-premises inference disclosed	No user count published; 450+ AI projects (Jul 2026)	Experimenting
National Grid	Alice Delahunty, Chief Technology and Innovation Officer, on the Group Executive Committee since Sep 2026	Bought: Vendor AI tools and Microsoft Copilot; internal GridGPT chat service (2024)	No LLM provider named; Microsoft Copilot and vendors' models	Microsoft Azure, where National Grid moved much of its IT estate and builds generative AI (Sep 2024); vendor SaaS such as NICE CXone	Not disclosed; AI scores 100% of New England contact-centre calls	Experimenting
NextEra Energy	EVP and Chief Risk Officer Petter Skantze, who oversees IT, NextEra Analytics and REWIRE	Hybrid: REWIRE and in-house optimizers, built with Google Cloud	Gemini Enterprise Agent Platform, TimesFM 2.5 and WeatherNext 2, plus proprietary algorithms	Google Cloud for REWIRE, GridComposer and Optos; no other cloud or company data center for AI workloads disclosed	Not disclosed; 100+ AI initiatives under execution (Oct 2026)	Experimenting
PG&E	EVP and CIO Ajay Waghray; SVP and Chief Data & Analytics Officer David Leach runs the central AI team	Hybrid: In-house grid models; vendor generative AI (AWS, Atomic Canyon, PolyAI)	In-house machine learning models, Atomic Canyon FERMI models; no general model vendor named	AWS for contact center AI; Neutron on on-site NVIDIA GPUs at Diablo Canyon; other hosting not disclosed	900 contact center agents and 1,300 Diablo Canyon workers; no company-wide count	Experimenting
Southern Company	EVP and CITO Hans Brown; Technology Organization moves under CLO Sterling Spainhour on Jan 1, 2027	Hybrid: Databricks data and AI platform; Copilot agents at Southern Nuclear; vendor AI in operations	Microsoft Copilot at Southern Nuclear; vendor models such as ThermalLimits.ai; other LLMs not disclosed	Microsoft's Copilot service at Southern Nuclear; Databricks on an unnamed cloud; no data center named; public AI tools barred for company data	850+ active Copilot users at Southern Nuclear (Dec 2025); no systemwide count	Experimenting

Questions

AI in utilities, answered

Which utilities are furthest along with AI?: In the Enterprise AI diffusion report, Dominion Energy, Duke Energy, Exelon, Iberdrola, National Grid, NextEra Energy, PG&E, and Southern Company are at Experimenting. A stage follows from the returns and rollouts a company has publicly reported in the last 18 months, not from its spending or plans.
How do utilities use AI?: Utilities use AI most in outage prediction and restoration, vegetation management, line and plant inspections, grid planning and contact centers. Exelon runs its POSEIDON restoration model at all 6 of its utilities (Dec 2025), National Grid uses satellite imagery and AI to cut tree-related outages in New England and scores every New England contact-centre call with AI (May 2026), Dominion Energy applies visual AI to drone imagery (Mar 2026), Duke Energy uses AWS-managed AI agents in interconnection studies (Sep 2026), and PG&E gives 900 contact center agents Agent Assist and 1,300 Diablo Canyon workers the Neutron document search. Each of these utilities is at Experimenting, because none has reported a return it credits to AI alone since May 2025.
What return have utilities reported from AI?: Utilities have reported no return to their own costs, revenue or headcount that they credit to AI alone since May 2025. Duke Energy says AI-driven optimization had saved customers approximately $500 million as of 2024 (Jul 2026), and NextEra Energy says GridComposer saved more than $20 million in 2026, money that went to customers (Sep 2026). PG&E credits $11.2 million saved from Jan 2025 to Jun 2026 to continuous monitoring that joins sensors with machine learning, Iberdrola credits a 75% cut in inspection response times and operating costs on Brazilian transmission lines to drones and AI together (Feb 2026), and NextEra Energy's about $150 million a year of REWIRE and AI savings is a 2026 to 2032 target, not a result.
Which AI platforms and models do utilities use?: Utilities mostly build AI with a major cloud provider and name few model vendors. NextEra Energy runs REWIRE on Google Cloud with Gemini Enterprise Agent Platform, TimesFM 2.5 and WeatherNext 2 (Dec 2025), Iberdrola uses Anthropic Claude and Amazon Nova through Amazon Bedrock (Aug 2024) alongside Microsoft Copilot, Duke Energy builds on AWS, Southern Company uses Databricks for data and AI (Feb 2026) and Microsoft Copilot at Southern Nuclear, and National Grid builds generative AI on Microsoft Azure (Sep 2024). Exelon runs its own SPARK platform and NVIDIA-based vision models on its drones, and PG&E runs Atomic Canyon's FERMI models on NVIDIA GPUs at Diablo Canyon.

Other sectors

Every company in the report