AI and EnergyWeekly briefing · Monday, October 19, 2026Covers Oct 12–18 · 10 stories
Grid operators move toward rules for large AI data centers. Ohio shows how much new power and transmission one project can require.
The energy sector is starting to treat large AI data centers as grid participants that must provide data and follow operating instructions. That raises the cost and complexity of building them, while making dedicated generation and transmission more likely.
Written with AI from the 10 stories we linked this week. The small numbers point to the story list at the end.
The week in brief
- Large AI loads move closer to mandatory grid rules.
- Ohio's proposed AI complex needs power on a utility scale.
- AI's water use increasingly extends beyond data center walls.
- Utilities are applying AI to leaks and regulatory research.
Large data centers move toward grid rules
NERC stakeholders approved proposed reliability standards for data centers, AI facilities, crypto miners, and other large computational loads. The rules next go to NERC's board and then, if adopted, to FERC; they would cover sites of at least 50 MW connected at 100 kV or above.12
The proposed rules would require covered facilities to give grid operators information, coordinate protection settings, and respond to operating instructions. They follow incidents in which data centers rapidly changed demand or dropped load, creating grid-stability risks.12
FERC auditors also found AI tools increasingly in use at registered power entities, alongside AI and registration risks in critical-infrastructure compliance audits.3
So whatThe largest AI facilities are becoming regulated electricity customers, not merely very large buyers of power. Their developers will face more engineering, data-sharing, and operating obligations, while grid operators gain more visibility and control.
Ohio puts a 10-GW AI project under a grid study
PJM filed an agreement to study SoftBank's proposed Flagship complex in Ohio under its state-agreement approach. The plan covers an additional 9.6 GW of data-center demand, 10.2 GW of gas generation, and three planned 765-kV transmission lines, with phases running through 2032.4
An initial 800-MW phase would use an AEP Ohio data-center tariff, while the broader project includes an estimated $33 billion for 9.2 GW of gas-fired generation.4
Separately, Scioto Analysis estimated that data-center growth could add $16 billion in statewide Ohio electricity costs from 2027 through 2034, but the estimate applies a national demand-and-price projection rather than an Ohio-specific grid simulation.5
So whatThe project points to a costly new pattern: very large AI campuses may need their own power plants and high-voltage lines. That can speed a project compared with waiting for general grid upgrades, but it ties AI growth more directly to gas construction and fights over who pays for shared network costs.
AI's water bill is largely paid at power plants
Researchers estimated that electricity generation could add 14 billion gallons a year to the water consumption associated with three planned large AI data centers in Michigan. Meta reported that indirect water use represented about 96% of its 2024 water consumption.6
Southern Water is trialing Lightsonics' system in Hampshire, which uses signals in Openreach fiber-optic cables and machine learning to identify likely pipe leaks. Lightsonics says an earlier trial across 650 kilometers of water network avoided losses of more than four million liters a day.7
Permeta launched a beta service that lets users query more than 500,000 FERC filings through Claude, ChatGPT, and Copilot, returning cited links to the underlying regulatory record.8
So whatData-center cooling is only a small part of AI's water problem where generation relies on water. At the same time, utilities can use AI and existing infrastructure to reduce losses and research costs, though those savings will not remove the resource pressure created by new electricity demand.
Caution: Southern Water's trial and Lightsonics' prior-trial result are company-reported7; Permeta's FERC.ai coverage and functionality are company claims8; AWS's claims about AI improving utility planning are from an AWS executive's opinion piece9.
This week’s stories
All 10 stories we linked in AI and Energy, Oct 12–18. Headlines open the original.
Discussed above
- 1Reliability standards for AI data centers, crypto miners move closer to adoption
- 2NERC Reliability Standards for Data Centers and Crypto Loads Pass Final Ballot, Advance to Board Review
- 3FERC Auditors Warn of AI, Registration Risks
- 4PJM plans to study SoftBank’s Ohio AI project under state agreement approach
- 5Study: Data Centers Could Add $16B to Ohio Electricity Costs
- 6AI's Water Strain Includes Surging Demand From Electricity Use
- 7Southern to look for leaks with fibre optics
- 8Tired of digging through FERC filings? Let AI do it for you
- 9Utilities can use AI to solve grid backlogs: AWS energy lead