Portland General Electric Says GridCARE Found 80+ MW for Data Centers

Portland General Electric said its AI work with GridCARE identified more than 80 MW of capacity for data center connections years earlier than first expected, Utility Dive reported on Oct. 22. US and European utilities are using AI mainly for load forecasting and maintenance, not to run the grid, while a National Grid Partners survey says nearly 80% of utility leaders have fully deployed at least one AI tool for large-load customer planning.

Executives told Utility Dive the sector is far from letting large language models act on physical infrastructure. Most deployments are narrow tools that support human decisions on load forecasting and vegetation management. Rick Lesley of QualityAI said many demonstrations perform well in controlled conditions, but evidence of consistent performance at scale remains limited.

The obstacle is audit, not capability. Outputs that are uneven or cannot be audited sit badly with strict safety and reliability rules. Guardrail Technologies, a vendor of AI security and governance software, reviewed federal filings and found that roughly 70% of energy and utility companies named AI as a cybersecurity risk. Only 23% of utilities and 9% of energy companies documented a process for managing it, according to that review.

What AI Means for Data Center Connections

The practical payoff so far is speed in finding room on the wire. Portland General Electric said its work with GridCARE, which models demand hourly to locate spare grid capacity, identified more than 80 MW for data center connections years earlier than first expected. That is one utility's statement relayed by Utility Dive, not an audited result, but it shows where AI is already touching interconnection timing.

Emerald AI, another National Grid Partners portfolio company, says its platform can cut a data center's power demand by up to 40% when the grid asks for curtailment, while keeping critical workloads running. That is a vendor claim. Voltgrid's earlier coverage of Google and Voltus expanding a PJM virtual power plant shows the same logic from the demand-response side. Pradeep Tagare, interim president of National Grid Partners, said machine-learning virtual power plants have run for about five years but are less automated than systems that independently control core grid operations.

Who Pays and Which Regulators Are Watching

The same National Grid Partners survey, covering the United States, the United Kingdom and Europe, found 87% of respondents saying outdated regulatory and rate case frameworks hinder returns on AI and innovation spending. It also found 83% saying the cost of building infrastructure for AI data center demand is being passed to residential customers through higher bills. The survey is run by the venture arm of National Grid, which invests in AI energy startups.

Regulators are starting to look. The Federal Energy Regulatory Commission released a cybersecurity audit of critical infrastructure entities earlier in October that noted increasing use of AI tools and warned their evolving capabilities may introduce access control risks. New York regulators recently directed utilities to file inventories of their AI use, and on Oct. 13 Connecticut Attorney General William Tong and Consumer Counsel Claire Coleman petitioned their state regulators to open a similar review.

What Is Still Unconfirmed About Utility AI

Nobody can say how much AI is already running inside utilities. Stephen Rose, chief executive of Render Networks, told Utility Dive it is still unclear how much AI is running and on what data, and he warned that utility records do not always match what is physically in the field. The 80 MW figure, the 40% curtailment figure and the cybersecurity percentages all come from companies or vendors with a stake in the technology. Tagare expects LLM-based grid management to gain momentum within a couple of years, but said agentic systems touching physical assets are not yet widely deployed.